Full text of the official published hearing record. Extracted from the source document — verify against the official record for citation.
- LIABILITY OR DENIABILITY? PLATFORM POWER AS SECTION 230 TURNS 30 [Senate Hearing 119-507] [From the U.S. Government Publishing Office] S. Hrg. 119-507 LIABILITY OR DENIABILITY? PLATFORM POWER AS SECTION 230 TURNS 30 ======================================================================= HEARING before the COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION UNITED STATES SENATE ONE HUNDRED NINETEENTH CONGRESS SECOND SESSION __________ MARCH 18, 2026 __________ Printed for the use of the Committee on Commerce, Science, and Transportation [GRAPHIC NOT AVAILABLE IN TIFF FORMAT] Available online: http://www.govinfo.gov ______ U.S. GOVERNMENT PUBLISHING OFFICE 64-395 PDF WASHINGTON : 2026 SENATE COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION ONE HUNDRED NINETEENTH CONGRESS SECOND SESSION TED CRUZ, Texas, Chairman JOHN THUNE, South Dakota MARIA CANTWELL, Washington, ROGER WICKER, Mississippi Ranking DEB FISCHER, Nebraska AMY KLOBUCHAR, Minnesota JERRY MORAN, Kansas BRIAN SCHATZ, Hawaii DAN SULLIVAN, Alaska EDWARD MARKEY, Massachusetts MARSHA BLACKBURN, Tennessee GARY PETERS, Michigan TODD YOUNG, Indiana TAMMY BALDWIN, Wisconsin TED BUDD, North Carolina TAMMY DUCKWORTH, Illinois ERIC SCHMITT, Missouri JACKY ROSEN, Nevada JOHN CURTIS, Utah BEN RAY LUJAN, New Mexico BERNIE MORENO, Ohio JOHN HICKENLOOPER, Colorado TIM SHEEHY, Montana JOHN FETTERMAN, Pennsylvania SHELLEY MOORE CAPITO, West Virginia ANDY KIM, New Jersey CYNTHIA LUMMIS, Wyoming LISA BLUNT ROCHESTER, Delaware Brad Grantz, Republican Staff Director Nicole Christus, Republican Deputy Staff Director Lila Harper Helms, Staff Director Melissa Porter, Deputy Staff Director C O N T E N T S ---------- Page Hearing held on March 18, 2026................................... 1 Statement of Senator Cruz........................................ 1 Statement of Senator Schatz...................................... 2 Statement of Senator Fischer..................................... 97 Statement of Senator Klobuchar................................... 98 Statement of Senator Schmitt..................................... 101 Statement of Senator Baldwin..................................... 103 Statement of Senator Curtis...................................... 105 Statement of Senator Lujan....................................... 107 Statement of Senator Capito...................................... 109 Statement of Senator Rosen....................................... 111 Statement of Senator Hickenlooper................................ 112 Statement of Senator Markey...................................... 114 Statement of Senator Blackburn................................... 116 Witnesses Daphne Keller, Director of Platform Regulation, Program in Law, Science, and Technology, Stanford Law School................... 4 Prepared statement........................................... 6 Nadine Farid Johnson, Policy Director, Knight First Amendment Institute...................................................... 9 Prepared statement........................................... 11 Matthew Bergman, Founding Attorney, Social Media Victims Law Center......................................................... 14 Prepared statement........................................... 15 Brad Carson, President, Americans for Responsible Innovation..... 88 Prepared statement........................................... 90 Appendix Hon. Ron Wyden, U.S. Senator from Oregon, prepared statement..... 119 Brandon Arnold, Executive Vice President, National Taxpayers Union, prepared statement...................................... 120 Laura Chadwick, President and CEO, Travel Technology Association, prepared statement............................................. 120 Letter dated March 17, 2026 to Hon. Ted Cruz and Hon. Maria Cantwell from Linda Moore, President and CEO, TechNet.......... 121 Letter dated March 17, 2026 to Hon. Ted Cruz and Hon. Maria Cantwell from Shae Gardner, Director of Policy and Research, LGBT Tech...................................................... 122 Letter dated March 17, 2026 to Hon. Ted Cruz and Hon. Maria Cantwell from Sarah O'Quinn, U.S. Director of Public Affairs, Center for Countering Digital Hate............................. 123 Letter dated March 17, 2026 to Hon. Ted Cruz and Hon. Maria Cantwell from Kinsey Fabrizio, President, Consumer Technology Association and Michael Petricone, Senior Vice President, Government Affairs, Consumer Technology Association............ 126 Letter dated March 18, 2026 to Hon. Ted Cruz and Hon. Maria Cantwell from Arthur D. Sidney. CEO and Principal, The Sidney Group.......................................................... 128 Response to written questions submitted to Daphne Keller by: Hon. Eric Schmitt............................................ 129 Hon. Maria Cantwell.......................................... 132 Hon. Amy Klobuchar........................................... 135 Hon. Brian Schatz............................................ 136 Hon. Edward Markey........................................... 138 Response to written questions submitted to Nadine Farid Johnson by: Hon. Maria Cantwell.......................................... 140 Hon. Amy Klobuchar........................................... 141 Hon. Brian Schatz............................................ 141 Response to written questions submitted to Matthew Bergman by: Hon. Amy Klobuchar........................................... 143 Hon. Brian Schatz............................................ 143 Hon. Ben Ray Lujan........................................... 146 Response to written questions submitted to Brad Carson by: Hon. Maria Cantwell.......................................... 163 Hon. Amy Klobuchar........................................... 166 Hon. Brian Schatz............................................ 167 LIABILITY OR DENIABILITY? PLATFORM POWER AS SECTION 230 TURNS 30 ---------- WEDNESDAY, MARCH 18, 2026 U.S. Senate, Committee on Commerce, Science, and Transportation, Washington, DC. The Committee met, pursuant to notice, at 10:11 a.m., in room SR-253, Russell Senate Office Building, Hon. Ted Cruz, Chairman of the Committee, presiding. Present: Senators Cruz [presiding], Fischer, Blackburn [presiding], Schmitt, Curtis, Moreno, Capito, Schatz, Klobuchar, Baldwin, Rosen, Lujan, Hickenlooper, and Markey. OPENING STATEMENT OF HON. TED CRUZ, U.S. SENATOR FROM TEXAS The Chairman. Good morning. The Senate Committee on Commerce, Science, and Transportation will come to order. Welcome to all the witnesses. Within our lifetimes, the Internet has impacted nearly every aspect of the world and our daily lives, especially how we communicate. It was only a short time ago that speech and newsworthiness was controlled by a handful of TV networks and giant newspaper publishers. If you held a position they didn't want to print or wasn't consistent with their political views, it didn't get said. The Internet changed that, allowing anyone to bypass these gatekeepers and shape public opinion with their own views. The Internet also created a new way to communicate anonymously and at greater scale through blogs, message boards, and comment sections. But with opportunity came legal questions. The law wasn't written for the internet's ease and anonymity. Holding a platform liable for the illegal speech of another person threatened potentially to overwhelm early Internet companies with ruinous lawsuits that would predictably result in less online speech. So Washington explicitly adopted a light touch regulatory approach with the enactment of the Telecommunications Act of 1996. Congress included Section 230 to ensure that online platforms would not be liable for the illegal speech of another person. It did so to preserve a competitive free market, and the text of Section 230 explicitly recognized that the Internet provided a ``forum for a true diversity of political discourse.'' But 30 years later, it seems that Big Tech has now become the new gatekeeper, the new speech police. If you disagree with a particular view, Big Tech doesn't answer that with more speech. They do not try to persuade. They do not debate. They simply make the view they disagree with disappear, and they silence you. That should scare everyone. What is even more concerning is how the Government hijacks Big Tech's powers to shape online discourse and to suppress dissenting views and undermine free speech. This isn't fiction. As I detailed in my report and hearings last year, the Biden Administration weaponized the Cybersecurity and Infrastructure Security Agency to bully Big Tech to censor lawful speech on COVID and on elections, disproportionately muzzling conservative voices. We should recognize and celebrate how the free market can cause a course correction against Big Tech censorship. Elon Musk's purchase of Twitter was one of the most important steps for free speech in decades. It showed that the censorship regime is not inevitable, and it can be challenged in the marketplace and shifted to allow the kind of diverse viewpoints that Section 230 envisioned. Congress must also consider every Constitutional tool we have to ensure and to prevent social media from harming Americans, especially children, while not incentivizing Big Tech censorship. The Take It Down Act, which I led together with Senator Klobuchar, demonstrates that Congress can pass targeted legislation to protect children and adults online. The law prohibits non-consensual intimate images, including such images created with artificial intelligence, and it creates a process to provide notice and takedown for victims, all without amending Section 230 or chilling lawful speech protected by the First Amendment. I have also introduced several other legislative reforms to actively support free speech online, including the Terms Act, which stops online platforms from weaponizing their terms of service to silence Americans and deny them access to essential products and services. And I will soon be introducing the Jawbone Act to stop Government agencies from bullying platforms into silencing the American people. The same reasons why Congress enacted Section 230, to prevent liability for a different person's speech, are still relevant. And I am concerned that a full repeal or sunset would lead platforms to engage in worse behavior, to engage in more censorship, to protect themselves from litigation. I also don't believe, as some of my colleagues have suggested, that we should use Section 230 reform to silence more lawful speech or to turn the Government into the arbiter. But we should consider whether reform of Section 230 is needed to encourage and to protect more speech online and to stop Big Tech censorship. No Government official, regardless of party, should have the power of censorship. I agree with John Stuart Mill that the best solution for bad ideas and for bad speech is better ideas and more speech. We don't need to use brute Government force, because the truth is much more powerful. I turn to Ranking Member Schatz. STATEMENT OF HON. BRIAN SCHATZ, U.S. SENATOR FROM HAWAII Senator Schatz. Thank you, Chairman Cruz, and thanks to our witnesses for being here today. It has been 30 years since Congress passed the Communications Decency Act, which included Section 230. But since then, everything has changed. In 1996, the Internet as we know it today was still taking shape. We were still using dial-up modems, and pagers, and paper maps. DVDs weren't even available yet, much less smartphones, social media, and AI. The world was a totally different place, and our laws reflected that. In the years since, we have all come to rely on the Internet in virtually every part of our lives. We have learned about its benefits and harms from personal experience. And we have had to keep up with every new iteration of ever-changing technology. And yet, even as people have adapted to the technology, tech companies have demanded that our laws don't change. There is this idea that Section 230 was somehow perfectly written 30 years ago, as if it is a Constitutional provision--as if it is the only Federal statute that must not be touched. That is preposterous. How we reform the law matters, of course, but to suggest that any reform would be catastrophic makes no sense. We are the Commerce Committee. We reform laws. Second, nobody thinks the internet, as it currently functions, is without problems. Whether you are a young person who has grown up with the Internet or you are an older user still figuring your way around, the Internet can often be a terrible place. It is chaotic, and depressing, and confusing. And given the chance, people would love something better. But really, the main reason to be having this conversation about Section 230 is that Congress can and should respond to changing circumstances. We do it all the time. For instance, we reauthorize the FAA every 5 years, even though the concept of flying has changed a lot less in the last 5 years than the Internet has changed in the last 30. Similarly, we amend and update laws to protect consumers from scams. For so long, tech companies have used Section 230 as an excuse to avoid taking meaningful action to protect users, but especially kids, from egregious harms, harassment and abuse, frauds and scams. It is not that they don't know what is happening or even why it is happening, it is that to do something about it would be to hurt their bottom line. And so long as Federal law provides a shield, why even bother? But this committee is not powerless here. We don't simply have to accept the terrible outcomes as a fact of modern life. We can work together and fix the law. I have a bipartisan bill called the Internet Pact Act. The Chairman, Senator Klobuchar, and others have amendments to Section 230. Section 230 is not one of the Ten Commandments. It is not a Constitutional provision. It is a Federal statute, and we are lawmakers. And this idea that we can't touch it, otherwise Internet freedom incinerates, is preposterous. And so, I am looking forward to a good, constructive, bipartisan discussion about what reforms are possible. Thank you. The Chairman. Thank you. And I would like to enthusiastically reiterate what Ranking Member Schatz just said, that Section 230 is not one of the Ten Commandments--or for fans of the history of the world, it is not one of the 15 commandments either. I would now like to introduce our witnesses for today. Our first witness is Daphne Keller. Ms. Keller is the Director of Platform Regulation at Stanford Law School's Program in Law, Science and Technology. She previously served as Associate General Counsel at Google, and her writing has been widely featured in both law journals and popular newspapers. Our second witness is Nadine Farid Johnson, Policy Director at the Knight First Amendment Institute. Ms. Farid Johnson was formerly an American Foreign Service Officer and today teaches at Columbia University School of International and Public Affairs. Our third witness is Matthew Bergman, Founding Attorney of the Social Media Victims Law Center. As an attorney, Mr. Bergman has worked extensively in product liability law, and his firm has filed lawsuits against several major social media companies, alleging addictive design flaws. Our final witness is Brad Carson, President and Co-Founder of Americans for Responsible Innovation. Mr. Carson served as the Representative for Oklahoma's 2nd Congressional District from 2001 to 2005 and was Acting Undersecretary of Defense for Personnel and Readiness under the Obama Administration. Ms. Keller, you are recognized for your opening statement. STATEMENT OF DAPHNE KELLER, DIRECTOR OF PLATFORM REGULATION, PROGRAM IN LAW, SCIENCE, AND TECHNOLOGY, STANFORD LAW SCHOOL Ms. Keller. Thank you very much. Chairman Cruz, Ranking Member Schatz, and members of the Committee, thank you for the opportunity to speak today. I am a lawyer with over 25 years of experience in practice and as an academic in the field of platform regulation, working in the U.S. and around the world. I am also a native Seattleite. I am, alas, not a native Hawaiian. Married to a Texan. And I firmly believe that some of the most fundamental issues before us today are nonpartisan, and I heard a lot of similar sentiments from the two of you, which I find encouraging. We are here to talk about a law that is deeply unpopular today, Section 230. But it may still be, as Churchill said of democracy, the worst option except for all the other ones. I agree, 230 is not sacrosanct. It is not even well written. But it happened to strike a balance that has served its purpose, and I do not see other options that I think would be an improvement so far. Any proposed legal change, no matter how well-intentioned, should be vetted for two things. First, is it Constitutional? Second, would it make things worse? I will speak to the Constitution first. The Internet is full of speech that most of us in this room would find offensive and likely consider harmful or dangerous, but a lot of that speech is also protected by the First Amendment, including a great deal of hate speech and mis- or disinformation. Lawmakers cannot tell platforms to remove or demote this speech, and eliminating Section 230 wouldn't change that or give them an obligation to remove the speech. The Constitution also limits important details about laws attempting to hold platforms liable for users' genuinely illegal speech--things like fraud, or defamation, or obscenity. In a mid-20th Century case, the Supreme Court rejected strict liability for booksellers because incentivizing them to purge their shelves would harm us, the reading public. It was not about the rights of the booksellers. It was about the analogs of those of us today who depend on the Internet to share our messages, to post our book reviews, to read restaurant reviews, to engage in political dissent. This is about our rights. Now, let's talk about the second constraint. Would changes to the law make things worse? This is a question about what would actually happen, what platforms and users, and importantly, governments would predictably do in a world without Section 230. That legal change would very likely make the Internet worse for users' speech rights without, I believe, actually making it any safer. And it would impose legal uncertainty and expense that today's incumbent giants could survive, but their smaller rivals could not. We have a lot of data to predict what happens when speech--sorry, when platforms are held liable for the speech of their users. Platforms receive huge numbers of false allegations under laws like the DMCA here or the Digital Services Act in Europe from people demanding removal of perfectly legal speech. Governments do this. Companies do this against their competitors. And platforms have strong incentives to simply comply. The idea that the First Amendment alone would cause platforms to stand up for their users' lawful speech when doing so is expensive and inconvenient is simply wrong. We can also make informed predictions about harms to users from things-- from bad content, from things like pro-eating disorder content or pro-suicide content. Platforms take down enormous quantities of this material now under their own voluntarily adopted rules. The point of Section 230 was to encourage them to do so and make sure that they could look at the content and make editorial decisions without risking liability. In an alternate timeline without Section 230, platforms that moderated content at all would have reason to purge anything remotely risky, and the other option would be to leave everything up and tolerate all kinds of harmful or illegal material. Finally, losing Section 230 would, I believe, be catastrophic for competition. I wrote in my testimony about VO, a platform that was very similar to YouTube, that got sued on similar grounds, won on similar grounds, but went bankrupt in the process. And so, now we don't have this, what could have been a major competitor to YouTube. In a 230 list world, we should expect years or decades of legal uncertainty and litigation expense, the kinds of things the giants can withstand, and their smaller rivals can't. None of this is to say that Congress's hands are tied. There are ways to make the Internet a better and safer place without setting new state imposed rules for speech. Robust Federal privacy protections would be a great start. My fellow witness has written about how middleware and interoperability could offer the exciting promise of a more diverse and competitive Internet in which speech preferences are controlled by users themselves. Section 230 can help make that a reality. Section 230 has proven value, and Congress should not abandon it for proposed solutions that have every chance of making things worse. I thank you for your time today and look forward to your questions. [The prepared statement of Ms. Keller follows:] Prepared Statement of Daphne Keller, Director of Platform Regulation, Program in Law, Science, and Technology, Stanford Law School Chairman Cruz, Ranking Member Cantwell, and Members of the Committee: Thank you for the opportunity to appear at this hearing and discuss the legal backbone of today's online speech environment: the law known as Section 230.\1\ Section 230 is widely maligned. But we should be realistic about the significant value it provides, and the harms that we would almost certainly face without it. --------------------------------------------------------------------------- \1\ As Blake Reid explains, the law is technically Section 230 of the Communications Act of 1934. --------------------------------------------------------------------------- I speak today based on twenty-five years of experience in platform regulation as both a practicing lawyer and a legal scholar. For ten years, I experienced the impact of platform laws firsthand as counsel for Google, including as legal lead for web search. For another ten years at Stanford, I have studied and written about the practical and policy alternatives to Section 230--including laws proposed or enacted around the world and here in the United States. Based on this experience, my take on Section 230 is much like Winston Churchill's take on democracy: Many forms of Government have been tried, and will be tried in this world of sin and woe. No one pretends that democracy is perfect or all-wise. Indeed it has been said that democracy is the worst form of Government except for all those other forms that have been tried from time to time[.] Churchill was addressing the UK Parliament in 1947, with the horrors of World War II and the Holocaust fresh in memory. He did not speak lightly of sin, woe, or the imperfection of human governance. We should not speak lightly of the very real dangers in the world today, or of the Internet's role in facilitating them. But Section 230, despite its flaws, has proven its value in achieving the very goals that will be discussed in this hearing. Experience here and in other countries illustrates the foreseeable damage threatened by many alternatives to Section 230. Congress could repeal or amend Section 230, but two major things would not change. First, the U.S. Constitution protects the vast majority of the hate speech, disinformation, and other offensive or dangerous ``lawful but awful'' speech online. Lawmakers cannot tell platforms to remove online speech if they have no constitutional power to restrain that speech in the first place. The idea that eliminating Section 230 would make platforms liable for users' constitutionally protected speech is simply false. Congress could write a law to punish platforms for carrying defamation, fraud, obscenity, or other truly unlawful speech. But if laws like that go too far in incentivizing intermediaries to suppress legal expression, they can also violate the First Amendment. This is the lesson of Smith v. California, a 1959 Supreme Court case. The law at issue in that case imposed liability only bookstores--but, the Court noted, a ``bookseller's self-censorship, compelled by the State, would be a censorship affecting the whole public, hardly less virulent for being privately administered.'' The whole public is affected by today's platform regulations, too. Those laws must respect ordinary Internet users' rights to receive and comment on news, share text messages and family vacation photos, read restaurant reviews on Yelp, and post book reviews on Amazon. The second constraint on lawmakers' options is practical. Well- intended regulations will succeed or fail based on what platforms and users actually do in response to changed legal rules. A law that aims to prevent violence but in fact incentivizes major platforms to prohibit all discussion of world events or politics is not a success-- particularly if the same speakers simply migrate to other platforms and foment even more violence. Neither is a law that aims to protect political speech but in practice would turn every platform into an identical, unrestricted cacophony. In the remainder of my testimony, I will discuss the important work that Section 230 is doing today, and question the likelihood that alternative proposals would yield better results. I will also distinguish viable approaches to regulating platforms' ``design'' from approaches that are constitutionally suspect. Section 230 is doing important work today Every platform speech regulation, including Section 230, represents a prediction--and a gamble--about the real-world behavior of platforms and Internet users. Section 230s prediction was based on what is now known as the ``moderator's dilemma.'' Its drafters were motivated by two court rulings. Together, the rulings told nascent Internet platforms that attempting to moderate content would put them in an editorial role, with legal responsibility for users' unlawful speech-- but that they could avoid this risk by refusing to intervene, and tolerating all manner of unlawful or harmful posts. Section 230 was designed to avoid the resulting perverse incentives. It sought to encourage platforms to create and enforce editorial policies, and protect them from the very real risk that doing so would lead to liability. Americans need not look far for evidence that Section 230s prediction was correct. The law has produced an ecosystem of small, medium, and large platforms that can all afford to exist, and to adopt diverse approaches to content moderation, because they can't easily be sued out of existence. Comparing claims that are not immunized by Section 230--federal crimes, intellectual property, trafficking, and prostitution--as well as foreign laws tells us a lot about the benefits Section 230 provides today. Encouraging moderation and avoiding the moderator's dilemma The experience of the video hosting platform Vimeo illustrates the moderator's dilemma in action. Vimeo employed content moderators in an effort to weed out both illegal uploaded content, such as obscenity, and content that violated the platform's own rules, such as hate speech. That choice to moderate is precisely what Section 230 encourages. But because copyright law does not offer the same immunity, Vimeo was drawn into litigation for well over a decade about whether those moderators might have seen and recognized, but failed to remove, copyright-infringing content. Vimeo has to date prevailed in court and survived the expense of litigation. Another smaller platform, Veoh, provides a more sobering example. Veoh and YouTube offered very similar services, and were sued on very similar copyright claims. Both ultimately won their cases, on nearly identical grounds. But being legally in the right was not enough to save Veoh. It went bankrupt in the process. YouTube, by contrast, was able to weather over $100 million in legal fees, and remains a behemoth today. Protection from devastating litigation costs, which accrue even in meritless lawsuits, is one of the most important benefits of Section 230.\2\ --------------------------------------------------------------------------- \2\ One study found that in 28 percent of cases in which platforms raised Section 230 defenses, courts found plaintiffs' claims invalid and resolved the cases without needing to consider Section 230. Courts relied on Section 230 as the primary basis for ruling in only 42 percent of cases. --------------------------------------------------------------------------- Without Section 230, most platforms would have two safe courses to avoid liability for claims like defamation. They could moderate so thoroughly that only the blandest and least controversial material remains; or they could avoid moderation entirely and leave users to face the resulting glut of scams, pornography, dangerous diet advice, advocacy of violence, and more. This would not be an overall safer or better Internet than the one we have now. Discouraging over-removal of lawful speech Abundant evidence shows that under the ``notice and takedown'' systems established by many non-230 platform liability laws, platforms' safest, cheapest, and easiest course is simply to honor every claim. This makes users' online expression vulnerable to a ``heckler's veto,'' and gives individuals, companies, and governments an avenue to silence speech simply by complaining about it. Even under the U.S. Digital Millennium Copyright Act--a law that attempts to protect speakers by allowing for appeals, reinstatement of lawful content, and penalties against bad faith takedown demands-- improper claims are extremely common, and far too often successful. Governments have used takedown demands to silence critical journalism and suppress video evidence of police brutality. Businesses have used them to target competitors. Activists and discredited scientists have used them to suppress inconvenient truths. Improper takedown claims are even more common under Europe's Right to Be Forgotten laws. Among claims targeting 7.8 million webpages for removal from search results, Google reports that nearly half were legally invalid. In theory, platforms should feel free to ignore complaints that target speech protected by the First Amendment. In practice, this is a pipe dream. Fighting is expensive, and platforms have little motivation to do it. They also often simply lack the information to know whether content is actually illegal. For example, they cannot know, and have little motivation to find out, whether allegedly defamatory news reporting about local political corruption is true or false. Even if platforms do have all the facts, the legality of speech often depends on complex doctrines like copyright fair use, or on nuanced, jurisdiction-specific precedent about who counts as a public figure in defamation cases. Litigating in the face of such uncertainty would be daunting even for more dedicated defenders of speech. That mix of bad platform incentives and legal uncertainty would define a world without Section 230. In such a world, we should expect platforms to regularly yield to takedown demands regardless of their merits. When platforms are forced to litigate, we should not expect a clear body of rules to emerge. There are simply too many varying legal questions to resolve. Even beyond the First Amendment issues, every state offers its own wide variety of tort claims, and every platform presents slightly or significantly different facts against which those claims may be tested. The room for plaintiffs to argue that a new case is not governed by precedent would be vast. Section 230 helps platforms resist improper pressure from the government, too. Without it, state and Federal officials could credibly threaten retaliation through, for example, targeted use of agencies' civil enforcement powers. Recent incidents in which broadcasters have yielded to what Chairman Cruz called ``mafioso'' tactics from FCC Chairman Brendan Carr illustrate the problem. Chairman Carr's threats temporarily drove comedian Jimmy Kimmel off the air. When CBS declined to air Steven Colbert's interview with Senate candidate James Talarico, the interview remained available on a Section-230-immunized platform, YouTube. The same dynamic could just as well arise with a liberal regulator and a conservative comedian or politician. As I tell my students, protecting lawful but unpopular speech from government ``jawboning'' is not a partisan issue. Everyone has reason to fear state power over their speech, whether under a current administration or a future one. History suggests that those who are societally marginalized or politically powerless are the likeliest victims of such abuse. It also suggests that when incumbents with close ties to and dependencies on government become too accommodating, it is the independent players who may show more spine. Section 230 is critical in allowing them to do so. Encouraging competition and diversity of online platforms A persistent myth in policy discussions holds that Section 230 only protects the biggest platforms. Nothing could be farther from the truth. Today's incumbents would survive the turbulence and uncertainty of a world without immunities. Their smaller competitors, like Veoh, very likely would not. This is the real backdrop when large incumbents ``come to the table'' and embrace changes to immunity. They are signaling that they can live with the consequences. If lawmakers want the Internet to evolve past its current state--if they want to constrain the power of today's tech giants--the law must enable new and niche competitors to thrive. Real-world defendants in Section 230 cases include newspapers, universities, libraries, employers, bloggers, and providers of spam protection and anti-fraud tools. They also include Internet infrastructure providers like domain name registrars. For the many smaller companies and non-profits the law protects, litigation costs remain daunting even with Section 230 in place. For early-stage startups, with a reported average $55,000 per month to cover all expenses, they may simply be insupportable. Section 230 gives the next generation of competitors a fighting chance against today's giants. Section 230, speech, and ``design'' A current trend in Section 230 litigation and legislation focuses on concepts like ``design'' and ``risk mitigation.'' Those words can mean many things. Some are perfectly fine and even promising. Others are in profound tension with both Section 230 and the First Amendment. In the ``perfectly fine'' category, design regulation can advance uncontroversial consumer protection or privacy goals. Limits on data collection and use generally do not implicate either Section 230 or the First Amendment. Neither do requirements to accurately disclose privacy practices or label options offered to users. More ambitiously, ``middleware'' and interoperability requirements could potentially give people more control over their online experiences--letting them choose among competing providers of content moderation without sacrificing the ``network effects'' of connections with friends and contacts. As I have argued, such an approach would avoid the major First Amendment problems that arise when lawmakers attempt to impose new, state-defined speech rules on Internet users. (Aggressive versions of middleware legislation would likely bring First Amendment objections from platforms, but those may not be insurmountable.) Remarkably, Section 230 itself anticipates and supports the development of such user controls. In the underappreciated subsection (c)(2)(B), the law immunizes providers of ``technical means to restrict access'' to unwanted material. In a far more problematic category, laws or claims that are nominally about design or data management can indirectly force platforms to change their rules for speech. Neither courts nor legislators may avoid the First Amendment by this sleight of hand. When California attempted to do that under its Age Appropriate Design Code, the Ninth Circuit correctly affirmed an injunction against relevant portions of the law, saying that they ``deputize[d] covered businesses into serving as censors for the State.'' The court reiterated its conclusion just last week. For similar reasons, lawmakers may not bypass the First Amendment by regulating algorithmic amplification, requiring platforms to promote or demote lawful speech based on state preference. In between these two extremes are a handful of features that are increasingly central in current litigation. These include autoplay for videos, infinite scroll, and notifications. Questions about whether and when such features legitimately lie outside of Section 230 or First Amendment protection are currently before courts. So, too, are questions about whether these features, divorced from the content they convey, violate any laws. Congress should not act prematurely, before this process plays out. Conclusion Section 230 has proven value. It gives platforms the freedom to set and enforce their own editorial policies, and protects them when they stand up against abusive takedown demands--including from the government. This is a difficult balance. Section 230 may not be, in Churchill's words, ``perfect or all-wise[.]'' But Congress should not abandon it for proposed solutions that have every chance of making things much worse. The Chairman. Thank you, Ms. Farid Johnson, you are recognized. STATEMENT OF NADINE FARID JOHNSON, POLICY DIRECTOR, KNIGHT FIRST AMENDMENT INSTITUTE Ms. Johnson. Thank you. Chairman Cruz, Ranking Member Schatz, distinguished members of this committee, thank you again for the opportunity to testify today. The issues before this committee today are immensely important ones. All of us here agree that the digital public sphere is not working for Americans or for our democracy, and the question is what to do about it. Section 230s protection is vital to free speech online, even if there are difficult questions about how far its protection should extend, questions that the courts are working through right now. While Section 230 should not be treated as sacrosanct, repealing it would do little to address the problems that all of us are most concerned about. And in some ways, it would make those problems worse. The better approach would be to pass structural regulation that would protect users' privacy, allow them to engage with platforms on their own terms or leave them more easily, and make the platforms more transparent and accountable to the public. If Congress is going to amend Section 230, it should make its protection conditional on platforms' compliance with transparency, privacy, and interoperability requirements. Section 230 effectively gave platforms the ability to moderate user content without having to fear that doing so would give rise to liability. With that protection, platforms now moderate content in many different ways, including by filtering or suppressing spam, pornography, and to varying degrees, hateful speech. While there is inevitably disagreement about what speech platforms should moderate and how, it is indisputable that the platforms would be unusable if they did not engage in moderation in ways that Section 230 was meant to protect. Some of the speech on the platforms is seriously harmful, denigrating, dangerous, polarizing. The Supreme Court has interpreted the First Amendment to deny to the Government the ability to punish hateful and dehumanizing speech or speech that promotes outrage. So it is important to recognize that the platforms would be protected for publishing most of this speech, even in the absence of Section 230, because the First Amendment would most likely be understood to protect it. Without Section 230, because the First Amendment protects harmful, outrages, and sensational speech, the platforms will not be required to remove it, but they would be motivated to take down speech that might plausibly give rise to liability, especially potentially defamatory speech. Rather than risk liability for such content, they would almost certainly remove it, including allegations that turn out to be true, thereby depriving users of access to socially valuable information critical to our public discourse. There is an opportunity now for Congress to act by putting forward legislation that is aligned with First Amendment rights and would foster a better online environment for people who use social media platforms, simultaneously minimizing the harms of social media while liberating users from the monopoly control over and suppression of their online speech. We have three proposals. First, lawmakers should establish legal protection for those who study the platforms in the public interest. Understanding the online experience of users, how algorithms target content, and how these decisions shape public discourse and ultimately our democracy is a necessary step toward informing the public about how the platforms operate. But today, those who studied the platforms are under the threat of serious legal liability. Passing a researcher safe harbor would improve public understanding of how the platforms are shaping society. One model for doing so is the Knight Institute's safe harbor proposal, a modified version of which was incorporated into the Platform Accountability and Transparency Act. Second, legislators can address several data privacy related issues that affects users' experience online. Platforms are successful in maintaining user engagement because they use the extensive information they gather about a user to recommend content. Legislators could require platforms to clearly inform users as to what data the platforms collect about them, how they use that data, and with whom they share it, and could also limit what information platforms collect. Congress can also pass legislation that prevents platforms from selling user information to data brokers. These privacy enhancing efforts would be a monumental step forward to protect all Americans, but especially minors, who engage online. Third, Congress can and should attack the platform's monopoly control over public discourse. The platforms have been extremely successful at maintaining control over the social media industry. In practice, this denies users meaningful choice over the platforms they use, and it makes it nearly impossible for new platforms to compete with the incumbents. The most direct way to address this problem would be to establish a requirement of interoperability, which would enable users to take their data and social networks with them when they leave a platform. We support these mandates as standalone bills. We also believe that conditioning Section 230 protections on these mandates can be done in a way that not only respects the limitations of the First Amendment, but actually promotes the values that underlie it. Thank you again for the opportunity to testify this morning. I look forward to your questions. [The prepared statement of Ms. Johnson follows:] Prepared Statement of Nadine Farid Johnson, Policy Director, Knight First Amendment Institute Chairman Cruz, Ranking Member Cantwell, and distinguished members of this Committee, thank you for the opportunity to testify today. My name is Nadine Farid Johnson, and I am the Policy Director at the Knight First Amendment Institute. The questions before this Committee today are immensely important ones: How can we protect and strengthen free speech rights even as technology evolves? How in this context can we protect other individual rights and democratic interests that are also fundamental? All of us here agree that the digital public sphere is not working for Americans or for our democracy. The question is what to do about it. As I will explain, Section 230s protection is vital to free speech online, even if there are difficult questions about how far its protection should extend--questions that the courts are working through right now. While Section 230 should not be treated as sacrosanct, repealing it would do little to address the problems that all of us are most concerned about, and in some ways it would make these problems worse. What I will explain today is that the better approach would be to pass structural regulation that would protect users' privacy, allow users to engage with platforms on their own terms or leave them more easily, and make the platforms more transparent and accountable to the public. If Congress is going to amend Section 230, what it should do is make Section 230s protection conditional on platforms' compliance with transparency, privacy, and interoperability requirements. Let me begin with the role that Section 230 plays in protecting free speech online. In the 1990s, when the Internet was in its infancy, two court decisions addressed whether efforts by platforms--in those days, bulletin board services--to curate their users' posts would subject the platforms to liability for any unlawful speech that they failed to take down. The decisions, Cubby v. CompuServe and Stratton Oakmont v. Prodigy, each assessed the potential liability online services might face for distributing user content. Taken together, those cases indicated that the more service providers curated the content posted on their platforms, the more likely they would be to face liability for that content. Had that regime of liability been allowed to go into full effect, it is easy to predict how the early platforms would have responded, because they would have had only two realistic options for avoiding crushing liability for the posts on their sites. They would have had to either allow everything to be published without moderation or aggressively take down any post that could conceivably provide a basis for liability. The first path would have turned the early Internet platforms into cesspools of pornography, hate, and spam; the latter would have turned them into sterile channels of play-speech, devoid of any allegations of wrongdoing--including truthful ones--by those with the means to sue. Section 230 effectively gave platforms the ability to moderate user content without having to fear that doing so would give rise to liability. The first part of the provision, codified in Section 230(c)(1), limited the liability of online service providers for content posted on their sites by third parties. The second, codified in section (c)(2), shielded online service providers for good faith efforts in content moderation. With the protection provided by Section 230, platforms now moderate content in many different ways, including in ways that are important to their users and to the public more broadly. The most popular social media platforms filter or suppress spam, pornography, and (to varying degrees) hateful speech directed at racial, religious, and other minorities. While there is inevitably disagreement about what speech platforms should moderate, and how they should moderate it, it is indisputable that the platforms would be unusable if they did not engage in moderation in ways that Section 230 was meant to protect. Although the very large online platforms dominate much of the conversation around Section 230, Section 230s protections reach the online experience far beyond social media. We often use ``platforms'' and ``social media'' interchangeably, but Section 230 applies to everything from news and entertainment aggregators, to community-driven encyclopedias and comment sites, to forums for artistic and cultural exchange. Section 230 protections have enabled features of our online lives that many of us take for granted, such as web search results. None of this is to deny that some of the speech on the platforms is seriously harmful. Some of these harms result from the publication of users' speech--the third-party content that platforms host--and some of these harms result from platform design. Some of the content that the platforms publish denigrates minorities; some of it is dangerous misinformation about public health; some of it is defamatory. Platforms' recommendation algorithms sometimes amplify this content; and sometimes they amplify other forms of polarizing content that are intended to generate engagement and outrage. It is important to recognize that the platforms would be protected for publishing most of this speech even in the absence of Section 230, because the First Amendment would most likely be understood to protect it. Even without Section 230, the platforms would not be liable for publishing the speech of their users that falls outside of the few, narrow exceptions to the First Amendment. Without Section 230, the platforms could perhaps be held liable for hosting defamation, or true threats, or incitement to imminent violence (at least so long as they had knowledge of the posts in question). But the Supreme Court has interpreted the First Amendment to deny to the government the ability to punish hateful and dehumanizing speech, or speech that promotes outrage. Even without Section 230 immunity, then, the platforms could rely directly on the First Amendment in responding to lawsuits over this kind of ``lawful, but awful'' speech. It is also important to recognize that Section 230 does not preclude platforms from being held liable for design decisions that are harmful--at least if those design decisions are not themselves expressive. The First Amendment has properly been understood to protect platforms' content-moderation decisions. Courts have, rightfully, upheld the platforms' First Amendment right to exercise traditional editorial functions--such as when they choose to feature or demote certain content, when they specify and enforce ``community standards'' that restrict what categories of content users can post, or when they attach warning labels to user content. That protection stems from a line of Supreme Court cases starting with Miami Herald Publishing Company v. Tornillo, which involved a statute concerning newspapers' editorial functions. Just two years ago, the Supreme Court made clear that this long line of cases applies to social media platforms when they ``decide which third-party content th[eir] feeds will display, or how the display will be ordered and organized.'' Moody v. NetChoice, LLC, 603 U.S. 707, 740 (2024). Neither Section 230 nor the First Amendment would protect platforms from liability, however, for design decisions that are not expressive and that cause harm. For example, in Lemmon v. Snap, Inc., bereaved parents sued the company behind the social media platform Snapchat for its allegedly negligent design of a ``Speed Filter'' that encouraged users to drive at excessive speeds. In reversing the district court's dismissal of the parents' complaint, the appellate court noted that a platform could still face liability for its provision of ``content- neutral tools'' where the liability stemmed from the platform's own acts--such as designing and making available to users the ``Speed Filter'' and corresponding reward system--and not the content posted by its users. There are of course hard questions around whether any particular design decision is expressive and for that reason implicates the First Amendment. Courts are addressing these questions in the product liability lawsuits recently filed against the social media platforms. Similar questions are also being adjudicated in the realm of generative artificial intelligence, specifically with respect to large language model-driven chatbots and minors' use of and engagement with those chatbots. As these and other relevant cases move forward, they may provide further insights into how Section 230 and the First Amendment apply in these contexts. If Congress eliminated Section 230, however, the main result would be to incentivize the platforms to take down speech that is socially valuable. As discussed above, because the First Amendment protects harmful, outrageous, and sensational speech, the platforms would not be required to remove it, even without Section 230. But they would be motivated to take down speech that might plausibly give rise to liability. That category would include, most significantly, potentially defamatory speech, such as speech that alleges wrongdoing by a specific individual. The First Amendment does not protect defamation. Truth is of course a defense to defamation liability, but the platforms do not have the ability to determine--definitively and at the scale that they operate--whether any particular allegation is true or not. This means that, absent Section 230, the platforms could face liability for leaving up such factual allegations about real people, and so they would almost certainly remove them, including allegations that turn out to be true. Needless to say, true allegations of wrongdoing-- particularly by government officials and powerful private actors--are socially valuable and important for public discourse. For all of these reasons, eliminating or substantially narrowing Section 230 has little demonstrated upside and dramatic potential downsides. Doing so would not address the harms that result from the platforms' publication of user speech, and it would create a significant incentive for the platforms to take down socially valuable speech. And even with Section 230 in place, the platforms are already potentially liable for harms that result from their own design decisions, at least to the extent that those design decisions are not protected by the First Amendment. If Congress is going to amend Section 230, what it should do is make Section 230s protection conditional on platforms' compliance with transparency, privacy, and interoperability requirements. These are challenging issues, but there is an opportunity now for Congress to act. Few believe that our system of free expression is working in a manner that best serves our democracy. Members of this Committee who want to address concerns about online harms should put forward legislation that is both aligned with First Amendment rights and would foster a better online environment for people who use social media platforms. Sunsetting Section 230 will not achieve that. The better course would be for Congress to enact structural changes that would better align the incentives of the large platforms with the interests of their users. Doing so holds the promise of simultaneously minimizing the harms of social media while liberating users from the monopoly control over--and suppression of--their online speech. Specifically, Congress should enact new mandates in transparency, privacy, data portability, and interoperability. First, lawmakers should establish legal protections for journalists and researchers who study the platforms in the public interest. Understanding the online experience of users--how algorithms target content, and how these decisions shape public discourse and, ultimately, our democracy--is a necessary step toward informing the public about how the platforms operate. These efforts are integral to holding the platforms accountable, and to ensuring that consumers have the information necessary to make informed choices about the social media networks they join. But today, researchers and journalists are stymied in their study of online platforms. The companies' terms of service generally ban the use of the tools necessary to study the platforms at scale, subjecting researchers to the threat of civil and potentially even criminal liability for public-interest investigations. Congress can and should address this problem by establishing a legal safe harbor for those who study the platforms in ethical and privacy- preserving ways to improve public understanding of how the platforms are shaping society. One model for doing so is the Knight Institute's safe-harbor proposal, a modified version of which was incorporated into the Platform Accountability and Transparency Act. Second, legislators can address several data privacy-related issues that affect users' experience online. Platforms are successful in maintaining user engagement because they use the extensive information they gather about a user--their viewing habits, interests, friends, likes, dislikes, beliefs--to recommend content accordingly. To facilitate transparency efforts, legislators could require platforms to clearly let users know what data the platforms collect about them, how the platforms use that data, and with whom they share that data. To protect users' privacy online, legislators can limit what information platforms collect. Prohibiting the collection of certain personal or behavioral data or circumscribing the uses of that data would limit the amount of information that is gathered by platforms and fed back into their algorithms, thereby lessening some of the algorithms' targeting ability. This would be a monumental step forward to protect all Americans but especially minors who engage online. And to further shield social media users from the misuse of their data, Congress can also pass legislation that prevents platforms from selling user information to data brokers, whose $270 billion-dollar business model involves the sale of Americans' sensitive data to businesses and governments. Third, Congress can and should attack the platforms' monopoly control over public discourse--directly. The platforms have been extremely successful at holding off their competitors and maintaining control over the social-media industry. They have been able to do so in large part because of the ``network effect''--the phenomenon by which the value of a product or service increases as more people use it. People want to be on the platforms their friends and families use, but once there, it becomes very difficult for users to leave, because the platforms do not allow them to take their social networks with them or to communicate with their old contacts from competitor services. This lock-in is anti-competitive, it denies users meaningful choice over the platforms they use, and it makes it nearly impossible for new platforms to compete with the incumbents. There are multiple ways to address this problem, but perhaps the most direct would be to establish a requirement of ``interoperability,'' which would enable users to take their data and social networks with them when they leave a platform. Competitors could be immunized for interoperating with other social media platforms, and large platforms could be required to facilitate this interoperation through Application Programming Interfaces (APIs). To be sure, any interoperability mandate would need to be carefully designed to avoid creating privacy risks, but this can and should be done. The Knight Institute notes that these mandates would benefit more Americans if enacted directly, so that they apply more broadly and even in situations and to actors not covered by or relying on Section 230. But we also recognize that there is logic in requiring actors that benefit from Section 230--especially platforms whose scale is made possible by its expansive immunity--to comply with a broader and stronger range of mandates meant to ensure that the platforms do not exploit that scale in ways that harm their users or public discourse. Conditioning Section 230 protections on things like researcher access, data privacy, and user controls can be done in a way that not only respects the limitations of the First Amendment, but actually promotes the values that underlie it. Whether these mandates are enacted directly or as a condition of Section 230 immunity, the opportunity exists to improve upon the current information landscape to protect the free flow of information while improving the user experience online. Mr. Chairman, Madam Ranking Member, and members of the Committee, I thank you again for the opportunity to testify this morning, and I look forward to your questions. The Chairman. Thank you. Mr. Bergman, you are recognized for your opening. STATEMENT OF MATTHEW BERGMAN, FOUNDING ATTORNEY, SOCIAL MEDIA VICTIMS LAW CENTER Mr. Bergman. Thank you. Chairman Cruz, Ranking Member Schatz, and esteemed members of this committee, it is an honor to be here to address the current state of Section 230. We appreciate the bipartisan leadership of this committee in addressing the carnage that is being inflicted on American young people through the deliberate design decisions of social media companies to target kids to enhance their profits over the safety of kids. And we commend this committee for its successful reporting out of the Kids Online Safety Act last session, which passed the Congress, and passed the Senate with 91 votes. As Justice Thomas observed, courts have interpreted Section 230 to confer sweeping immunity on some of the largest companies in the world. And in order to convey the human cost of this expansive definition, I have asked three families who have been directly affected by Section 230 to come before this committee and for you to hear their stories. Sitting to my left is Tammy Rodriguez of Enfield, Connecticut. She is the mother of Selena Rodriguez, forever 11. In the words of her sister, Destiny, ``Selena could light up a room as soon as she walked in. And you knew the second she walked in because the light came--with the light came a lot of noise''. Selena was so addicted to social media that she became physically violent when there was any effort to turn off the machine or take it away. As depicted in the written testimony, she was targeted through the deliberate design decisions of these companies with online sexual abusers and predators, and you can see the conversations that they had with her that would make any parent blanch. She developed such despair and depression that she took her life on Snapchat and recorded a song that was provided to her on Snapchat, Forever 11. Tony and Brandy Roberts are from New Liberia, Louisiana. They are the parents of England Roberts, age 14, forever. England was the youngest of five children. She was vivacious, charismatic, and sassy. In the words of her father, ``Tony, she made every day seem like Christmas''. Tony thought that he was monitoring his daughter's online content. He just didn't know how to do it. She was being pulled in a dark world, and as you can see in the written materials, developed more and more suicidal ideation based on the deliberate design decisions. And additionally, in the materials that we have submitted to this committee, you can see a chart showing on a daily basis how much she was being targeted after hours when her parents thought she was safe in her home and sleep. But she took her life by mimicking a depiction of suicide online that was targeted to her. And as recently as this morning, that material is still on Instagram. Jenny Cenario is the mother of Mason Edens, forever 16. In the words of his stepdad, ``Mason was an all-American football player, but there was a whole other side to him, with a gentle, loving, positive energy about him''. As a young high school boy, he broke up with his girlfriend and sought affirming content on TikTok to mend his broken heart. And instead, without asking for it, he was targeted with suicidal videos, encouraging him to take his life with a shotgun, which is what he did in his parents' home. And I have included clips to those things that TikTok targeted him with, and I would ask every parent in this room to look at those videos and ask whether a company that targets children with this content should be accorded unfettered immunity from basic principles of liability and human decency. Every parent in this room needs--these cases have nothing to do with protecting speech. They are about the deliberate design decisions of companies to prioritize profits over the lives and safety of their children. Yet in every one of these cases, social media has tried to dismiss these cases based on Section 230. Section 230, as written, sought to maximize user control over what information is received by individuals. The statute explicitly sought to empower parents and embolden law enforcement. And yet, this original intent has been thrown by the wayside by these interpretations. In the words of Justice Thomas, ``in the platform's world, they are fully responsible for their websites when it results in Constitutional protestations, but the moment that responsibility could lead to liability, they can dis-blame any obligation and enjoy greater protections from suit than nearly under any other industry''. We ask that this committee reform Section 230 to conform to its original intent to require that companies follow--that tech companies follow the rules that every other company in America follows, and that the duty of reasonable care not be a duty that is immune from only one segment of our economy. Thank you. [The prepared statement of Mr. Bergman follows:] Prepared Statement of Matthew Bergman, Founding Attorney, Social Media Victims Law Center (SMVLC) Chairman Cruz and Ranking Member Cantwell, I am honored to appear before this Committee with these distinguished panelists to discuss the relationship between Section 230 of the Communications Decency Act and the youth mental health and safety crisis that is currently ravaging our country. I look forward to discussing legislative reforms to restore the important statute to its laudatory purpose and to better protect America's kids and teens. Under your bipartisan leadership, this Committee has been at the forefront in changing the national dialog on social media, youth mental health, and the quest for online accountability. Since 2022, I have had the privilege of working with Republican and Democratic members of this Committee and their staffs on seeking legislative solutions to ameliorate the clear and present danger that social media poses to American's children. Every time I'm invited to discuss these issues there is one fundamental question that arises: how did we get here? How are technology companies able to design apps with such an indisputably corrosive impact on the mental and physical health of kids, earn billions of dollars in profit, and evade legal accountability? The answer boils down to two words: Section 230. In considering legislative solutions to better protect the public from unreasonably dangerous social media apps, it is therefore appropriate that 30 years after its enactment, this Committee take a fresh look at Section 230 and evaluate where the statute has been effective in promoting commerce and the free exchange of ideas and where reforms are desperately needed to incentivize companies to prevent harm to kids and teens before parents have to suffer immeasurably through the unimaginable consequences of congressional inaction. I have asked three parents who have been deeply impacted by the Section 230s expansive interpretation to join me in the hearing room today. Tammy Rodriguez from Enfield, Massachusetts is the mother of Selena Rodriguez who died of suicide in June 2021 at the age of 11 after being sexually abused on Instagram and TikTok. In January 2022, Tammy filed the first product liability claim against social media company involving an injured child. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Toney and Brandy Roberts live in New Iberia, Louisiana and are the parents of Englyn Roberts who took her life in 2020 after becoming addicted to social media and receiving unsolicited suicidal material in Instagram and TikTok. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Jennie DeSerio, from Centerton, Arkansas, is the mother of Mason Edens who shot himself to death in November 2022 at the age of 16. Mason had sought out affirming content after breaking up with his high school girlfriend but instead was deluged with a steady stream suicidal videos promoting suicide by shotgun. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Later in my testimony, I will provide more details on these tragic and several other cases in the context of Section 230. I believe it is imperative that any discussion of Section 230 reform focus on the individuals most impacted by legislative decision making: our children. I hope that my experience as a legal advocate for families whose children were injured or killed because of the deliberate decisions of social media companies to prioritize profits over the lives of their children will provide a helpful perspective to this Committee as it undertakes this critical task. A. Background As someone who was dragged kicking and screaming into the computer age, I am an unlikely candidate to opine on the intersection of law and technology. However, in retrospect, it cannot be mere coincidence that my first legal job was working for Ninth Circuit Court of Appeals Judge Diarmuid O'Scannlain, author of the Barnes v. Yahoo decision, which set forth the boundaries of Section 230 immunity. I parlayed this experience into a two-year judicial clerkship for Judge Bobby Baldock on the Tenth Circuit Court of Appeals before commencing my legal practice with a large national law firm repressuring companies facing environmental liabilities. I spent the next 25 years representing veterans and retired construction workers suffering from lung cancers and other illnesses caused by unnecessary toxic exposures in product liability actions throughout the Pacific Northwest. My career trajectory changed in the fall of 2021, after Frances Haugen's heroic and historic testimony before this Committee revealed that social media companies had been deliberately disregarding the severe mental and physical health harms that their apps have been inflicting on American kids for years. The leaked documents disclosed in the Wall Street Journal and Washington Post revealed a level of outrageous corporate misconduct that made every case I'd previously worked on pale by comparison. Yet virtually every effort to hold social media companies legally accountable had been thwarted through Section 230s misapplication, with most actions not even allowed to proceed past the pleading stage. After reading the Ninth Circuit's decision in Lemmon v. Snap, my colleagues and I charted a very narrow legal theory that might legally permit certain cases brought by parents to proceed despite Section 230. It was then that I formed the Social Media Victims Law Center in November 2021 to try against all odds to hold social media companies accountable for the veritable carnage their platforms are inflicting on America's kids. Currently, we are the only law firm exclusively focused on representing children and young adults injured or killed by social media apps or AI chatbots. In the ensuing 3.5 years, I have met with hundreds of parents who have lost children to suicide, accidental death, and drug overdose instigated by defective social media apps; children who almost died from eating disorders after being deluged with algorithmic designs meant to target and manipulate young brains with anorexic content; kids who have suffered both physical and online sexual abuse from adult predators facilitated by social media design; and children who have been treated for severe depression, anxiety and body dysmorphia because of cold, calculated decisions made by social media companies to keep these kids online no matter what the human cost. I am incredibly privileged to know and represent these families, and I'm constantly in awe of their passion and dedication. They are selflessly dedicated to ensuring that no other family must live through their nightmare of burying their own children. On their behalf and inspired by their steadfast grace in the face of unimaginable pain, I beseech you to take action to reform Section 230 so that that attorneys like me are no longer needed because harm is stopped before it can occur. These parents don't want more lawsuits, they want our kids to be happy, healthy, and safe online. B. Section 230 and its Critics ``The twenty-six words that created the Internet.'' In the three decades since its passage, Section 230 of the Communications Decency Act has achieved a degree of legend that far outpaces both the plaint text of the law itself and the express purposes underpinning it. The law mandates that ``[n]o provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider,'' 47 U.S.C. Sec. 230(c), and requires that ``[n]o cause of action may be brought and no liability may be imposed under any State or local law that is inconsistent with this section.'' 47 U.S.C. Sec. 230(e)(3). At the same time, Section 230 explicitly provides that ``[n]othing in this section shall be construed to prevent any State from enforcing any State law that is consistent with this section.'' 47 U.S.C. Sec. 230(e)(3). There are no references to sweeping immunity anywhere in the text. Instead, the current dismal state of the law is owed almost entirely to judicial interpretation. The Fourth Circuit in Zeran v. America Online, Inc. was the first court to suggest any form of Federal immunity under the ``plain language'' of Section 230. Ironically, the court cloaked its analysis in the language of free speech: The purpose of this statutory immunity is not difficult to discern. Congress recognized the threat that tort-based lawsuits pose to freedom of speech in the new and burgeoning Internet medium. The imposition of tort liability on service providers for the communications of others represented, for Congress, simply another form of intrusive government regulation of speech. Section 230 was enacted, in part, to maintain the robust nature of Internet communication and, accordingly, to keep government interference in the medium to a minimum. In specific statutory findings, Congress recognized the Internet and interactive computer services as offering ``a forum for a true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity.'' It also found that the Internet and interactive computer services ``have flourished, to the benefit of all Americans, with a minimum of government regulation.'' Congress further stated that it is ``the policy of the United States . . . to preserve the vibrant and competitive free market that presently exists for the Internet and other interactive computer services, unfettered by Federal or State regulation.'' Zeran v. Am. Online, Inc., 129 F.3d 327, 330 (4th Cir. 1997). Despite these laudable ideals, courts have expanded upon the Zeran analysis throughout the years, to the point that Section 230 has become utterly unrecognizable to either the text or the context in which it was passed. And predictably, tech companies have seized upon this trend to evade all possible legal accountability simply because third-party content is found somewhere in the causal chain of their misconduct. In Barnes v. Yahoo!, Inc., the Ninth Circuit stated that ``what matters is not the name of the cause of action . . . [but] whether the cause of action inherently requires the court to treat the defendant as the `publisher or speaker' of content provided by another.'' From this, social media companies insist that because everything they do involves third-party content, then every cause of action must necessarily treat them as publishers or speakers. In M.P. by and through Pinckney v. Meta Platforms Inc., the Fourth Circuit held that claims of products liability and negligence were barred by Section 230 because the claims at issue were ``inextricably intertwined with Facebook's role as a publisher of third-party content'' and that ``a newspaper company does not cease to be a publisher simply because it prioritizes engagement in sorting its content.'' From this, social media companies insist that every type of products liability and negligence claim must necessarily be foreclosed by Section 230, even where the basis arises from targeting children with deliberately addictive design features to prioritize engagement over user safety. And in Force v. Facebook, Inc., a split majority of the Second Circuit held that ``the text of Section 230 should be construed broadly in favor of immunity'' and that the use of algorithms for ``arranging and distributing third-party information inherently forms `connections' and `matches' among speakers, content, and viewers of content, whether in interactive Internet forums or in more traditional media.'' From this, social media companies insist that every algorithmic design must necessarily be exempt from liability no matter how they are designed or optimized. That said, there is a growing chorus among appellate courts pushing back on Zeran's original sin of reading Section 230 far beyond its plain text. Chief Judge Katzmann's partial dissent in Force is widely cited as the more persuasive interpretation of Section 230 and a better analytical framework for adding additional free speech priorities while holding companies accountable for weaponizing platform design features. Judge Katzmann recognized that the expansion of Section 230, ``a provision that was designed to encourage computer service providers to shield minors from obscene material,'' was instead being used to immunize companies that target and prey on vulnerable children. Instead, Judge Katzmann emphasized the law does not grant publishers Section 230 immunity ``for the full range of activities in which they might engage,'' only for claims faulting the defendant corporation for ``deciding whether to publish, withdraw, postpone or alter content.'' ``As is so often the case with new technologies, the very qualities that drive social media's success--its ease of use, open access, and ability to connect the world--have also spawned its demons.'' Judge Katzmann believed that ``claims regarding the algorithms are a poor fit for statutory immunity'' because social media companies could ``stop using the algorithms altogether'' or ``modify its algorithms'' to remove the dangerous defects in their design--all without modifying or removing a single piece of content. Considering the ways in which social media companies, with judicial approval, had manipulated the Communications Decency Act to avoid any accountability, Judge Katzmann suggested that Congress ``revisit the CDA to better calibrate the circumstances where such immunization is appropriate and inappropriate in light of congressional purposes.'' In the seven years since Force, Congress has declined to take Judge Katzmann up on his suggestion. However, jurists have increasingly begun to resist the allure of Zeran's overreach and embrace Judge Katzmann's more restrained reading of Section 230. Most notably, Justice Thomas has repeatedly castigated the judicial expansion of Section 230s plain text. In Malwarebytes, Inc. v. Enigma software Group USA, LLC, Justice Thomas' statement upon the denial of a writ of certiorari recognized that when Congress first enacted Section 230, ``most of today's major Internet platforms did not exist,'' and that since that time many courts have ``construed the law broadly to confer sweeping immunity on some of the largest companies in the world.'' Indeed, Justice Thomas decried contemporary interpretations that would permit a company to ``solicit thousands of potentially defamatory statements, `selec[t] and edi[t] . . . for publication' several of those statements, add commentary, and then feature the final product prominently over other submissions--all while enjoying immunity.'' Paring back this sweeping immunity, Justice Thomas explained, ``would not necessarily render defendants liable for online misconduct. It simply would give plaintiffs a chance to raise their claims in the first place.'' In 2024, Justice Thomas again sounded the alarm, again in the context of the high court declining to take up review of Section 230. Dissenting from the denial of certiorari in Doe Through Roe v. Snap, Inc., a case where Section 230 was held to immunize online platforms for facilitating child sex trafficking, Justice Thomas urged the Court to at least ``address whether social-media platforms--some of the largest and most powerful companies in the world--can be held responsible for their own misconduct.'' Moreover, Justice Thomas recognized that ``there is danger in delay.'' He articulated, in stark terms, how social media companies alternately rely upon either the First Amendment or Section 230 to evade accountability: ``In the platforms' world, they are fully responsible for their websites when it results in constitutional protections, but the moment that responsibility could lead to liability, they can disclaim any obligations and enjoy greater protections from suit than nearly any other industry.'' Other courts have begun follow Justice Thomas' criticisms. In A.B. v. Salesforce, Inc., the Fifth Circuit agreed that the law does not mandate immunity ``solely because a cause of action would not otherwise have accrued but for the third-party content.'' There, the court that companies may be held liable for conduct `` `that is properly attributable to them,' even if third-party speech exists somewhere upstream.'' In Anderson v. TikTok, Inc., Judge Matey of the Third Circuit remarked that Section 230 ``rides in to rescue corporations from virtually any claim loosely related to content posted by a third party, no matter the cause of action and whatever the provider's actions.'' The result, of course, is that current judicial interpretations of Section 230 ``immunizes platforms from the consequences of their own conduct and permits platforms to ignore the ordinary obligation'' of reasonable care and safe product design. Additionally, in Calise v. Meta Platforms, Inc., the Ninth Circuit found that claims based on fraudulent third-party advertisements posted in violation of Meta's terms of service were not barred by Section 230, holding that ``it is not enough that a claim, including its underlying facts, stems from third-party content for Sec. 230 immunity to apply.'' Id. at 742. C. How Social Media Hurts Kids In order to consider the real-world impact of the courts' overly expansive interpretation of Section 230, it is essential to understand the causal nexus between social media company's deliberate design decisions and the current mental health crisis plaguing America's kids. I am therefore providing the Committee with a synopsis of what we have learned from national experts on child neurology and development, internal company documents and testimony from company personal in the four years since we began litigating with the social media giants. Shockingly, the most compelling evidence on the causal on the relationship between social media and youth mental health is contained in social media companies' internal documents made public for the first time in the trial completed last week in Los Angeles. In 2023, Surgeon General Vivek Murthy issued a stark advisory on youth mental health, observing that, ``[t]he great irony of social media is that the more you immerse yourself in it, the lonelier and more depressed you become.'' Pre-teens and teens are particularly vulnerable to problematic use of social media and the resulting negative health outcomes. Social media causes or contributes to mental health harms such as addiction, problematic usage, anxiety, depression, body dysmorphia, eating disorders, poor sleep, suicide, and self- injury. The paths by which media use in general, and social media use in particular, are related to these mental health outcomes are complex and inter-related. D. The Psychology of ``Flow'' Humans bring innate and acquired skills to the challenges they face. When skills are high and challenges are low, the task at hand is sufficiently easy that it can induce boredom. Conversely, when challenges are high and skills are low, the task is sufficiently hard that it can induce anxiety. The psychologist Mihaly Csikszentmihalyi introduced the concept of ``flow'' as that mental state where the challenges and skills are sufficiently balanced that the experience is engrossing, engagement is easy, enjoyment is high, and time passes effortlessly. Flow states result in dopamine release and are inherently pleasurable. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In the ``real world,'' flow can be achieved by such things as well- timed promotions at work so that one feels deployed at the limit, but not beyond, their skill set, or in gaming contexts such as chess clubs by finding players that are worthy opponents. Both of those examples take time and effort and maybe even luck to achieve. Job promotions are rarely ``perfectly timed'' if they happen at all, and finding the ``right'' chess player can be challenging. With effort, some people can find flow in art, sports, music, and even work. Although Csikszentmihalyi maintained that ``flow'' was the key to a happy and fulfilling life, he cautioned that is not inherently or universally a ``good'' thing. He argued it could be misused in business and war, and that ``mountaineers'' and ``gamblers'' could become obsessed with it to the point of neglecting other aspects of their lives. Structured by algorithms as opposed to real world constraints, the online experience can create flow instantaneously and effortlessly and maintain it indefinitely. Consider the simple game of ``Candy Crush,'' which launched in 2012 and continues to command the attention of hundreds of millions of players. It requires no skill to play for the first time and so anxiety is low. In fact, reading the rules is unnecessary. The game is intuitive and there is no barrier--no friction--to beginning. Once a player begins playing, the game quickly and seamlessly ratchets up its difficulty in accordance with a player's skill and engagement: never so hard that it becomes frustrating, never so easy that it becomes boring. Two players starting at the same time play different games, but each is likely to find theirs enjoyable. Keeping people in a flow state is an engagement strategy that many social media sites actively deploy. People in a state of flow are, by definition, deeply engaged in the experience and less mindful of outside distractions or perturbations: it is an ``escape.'' This explains in part why people with underlying mental health conditions, or particularly disturbing realities (e.g., challenging socioeconomic circumstances), are more susceptible to the allure of a flow state and at greater risk for becoming addicted to what provides it to them--whether this is alcohol, drugs, gambling, or social media. To that end, Yao Qin and other researchers specifically investigated if flow (which they parameterized as enjoyment, concentration and time distortion) was associated with problematic TikTok use and found a strong correlation. E. Pre-Teen and Teen Brain Development Pre-teens and teens are particularly vulnerable to mental health harms from social media use due to their biological and psychological development. Children are not simply small adults, and the impact of any experience or exposure must be understood in the context of their ongoing biological and psychological development. Because of their brain development, they are particularly vulnerable to experiences such as ``FOMO'' (``fear of missing out'') and social contagion. Human brains develop throughout adolescence into early adulthood. Different regions mature at different rates. At birth, the brainstem and cerebellum are highly developed, supporting vital functions such as breathing, heart rate, and basic motor control. During infancy and early childhood, the limbic system, particularly the amygdala and hippocampus, rapidly develops, facilitating emotional responses and memory formation. Later, the cerebral cortex, responsible for higher cognitive functions, undergoes significant growth, with sensory and motor areas maturing first, followed by language centers, which develop rapidly in early childhood. As demonstrated in the attached illustration from Meta's internal documents furnished to this Committee by Frances Haugen, the prefrontal cortex, often referred to as the ``CEO of the brain'' because it is essential for decision-making, impulse control, and complex reasoning, is the last to fully mature, and typically completes development at around age 26. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] During adolescence, synaptic pruning strengthens important neural connections while eliminating less commonly used ones, refining cognitive abilities. Myelination, the process of insulating nerve fibers to improve communication between brain regions, progresses throughout childhood and adolescence, with the prefrontal cortex again being the last region to complete the process. This prolonged development explains why teens and emerging adults may struggle with long-term planning and impulse control compared to fully mature adults, and why social media can have significant impacts on children and adolescents. The figure below from the University of Cambridge shows when different brain regions reach 100 percent capacity. White matter volume (shown in dark blue) is the part of the brain that plays a crucial role in memory, attention, and decision making. The blue vertical line has been added to age 13, corresponding with the current ``minimal'' age for social media usage set by industry: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] ``Executive Function'' is a foundational construct used by cognitive psychology and neuroscience to describe a set of skills that emerge as the brain develops. It is comprised of several key capacities. All these play essential roles in human development and function and underpin both reactions to and effects of environmental stimuli. Children's developmental trajectories are highly individualized, which is to say there is considerable variability in the age at which these capacities are fully present. When we say that the ``typical'' 5-year-old can do X it means that as many as 1/2 to 1/3 cannot do it yet. Further, cognitive capacity is modulated by inhibitory control or ``self-monitoring.'' What this means is that even if a child ``knows'' the right thing to choose or to do, their lack of impulse control might make them get it wrong or not act appropriately. Optimal executive involves a titration of latency (delay) and accuracy, and executive function is essential to sound decision making. This lack of fully developed executive functioning makes adolescents particularly vulnerable to the harmful effects of social media. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This is something that the platforms appear to realize. For example, TikTok found that even among those that enabled their screen management tools, they saw no benefits (i.e., screen time use was not reduced) for their <18-year-old users, which they explained by saying ``minors do not have the executive function to control their screen time.'' F. Addiction Addiction is a based on directly measurable psychological and physiological attributes related to reliance on and/or withdrawal from a substrate. It is a complex condition characterized by compulsive engagement in rewarding stimuli despite adverse consequences. It often involves substances like drugs or alcohol. Teenagers and young adults who biologically lack higher cortical functioning including impulse control are more vulnerable to potentially addictive substances and behaviors. Epidemiological studies have shown that earlier onset of drug intake is associated with greater likelihood of development of substance use problems. In fact, the majority of problematic substance users (e.g., tobacco and alcohol) begin usage before the age of 21. Addiction can also be due to tolerance and withdrawal from certain behaviors. Not all behavioral addictions are currently recognized in the Diagnostic and Statistical Manual of Mental Disorder, Fifth Edition, (DSM-5) which is the American Psychiatric Association's (APA) guide to mental and brain-related conditions. However, there is an increasing recognition of the need for the DSM-5 to do so. To date, only a single behavioral addiction, gambling, is officially recognized by the DSM-5. The DSM-5 considered including ``gaming disorder'' in its 2013 edition but determined it was ``in need of further study.'' Meta has rightly recognized in an internal document that ``medical diagnoses change definitions over time based on new evidence.'' In 2022, the World Health Organization (WHO) did recognize the existence of ``gaming disorder'' as a clinical entity, and it is included as a diagnosis in the International Classifications of Disease 11 (ICD-11) which is the manual that physicians use to diagnose patients. Recently, the American Psychiatric Association has recognized ``technology addiction'' as ``excessive and compulsive use of the Internet or online activities [that] can lead to negative consequences in various aspects of an individual's life.'' Social media addiction is recognized as its own condition, characterized as ``involv[ing] problematic and compulsive use of social media; an obsessive need to check and update social media platforms, often resulting in problems in functioning and disrupted real-world relationships.'' The APA further recognizes that ``children and adolescents are particularly vulnerable to technological addiction because their brains are still developing'' and ``excessive problematic use of social media'' has the potential to develop into a behavioral addiction for children and adolescents. Because of the increased risks of social media to youth, several national associations and reports have been published with recommendations for actions that can be taken to help reduce the risk of mental health injury. Some of these reports include recommendations made in the textbook, Handbook of Children and Screens, as well as the ``Social Media and Youth Mental Health'' Report by the U.S. Surgeon General in 2023; ``Social Media and Adolescent Health'' by the National Academies of Sciences, Engineering, and Medicine in 2024; ``Health Advisory on Social Media Use in Adolescence'' by the American Psychological Association in 202358; and a report by the Jed Foundation in 2024. The American Academy of Pediatrics also recognizes problematic use and social media addiction and advises parents that ``[i]t's also important to recognize that it's not something wrong with the teen using the platform causing them to feel this way; many interactive technologies are specifically designed to capture and hold a user's interest. It can be hard for children and teens to overcome those design features.'' The National Eating Disorders Association also recognizes that ``research is increasingly clear that media does indeed contribute and that exposure to and pressure exerted by media increase body dissatisfaction and disordered eating.'' Taken together, these consensus statements leave little doubt that leading professional medical and psychological organizations recognize that social media and its problematic and addictive usage is harming children and teenagers today. G. Validated Social Media Addiction Scales A variety of screening instruments for what has been called ``Social Network Use Disorder'' have been developed. A systematic review of the scales revealed that two of them have the best validation data to support them: the Social Media Disorder Scale (SMDS) and the Bergen Social Media Addiction Scale Short Form (BSMAS-SF). Both measure features of substance abuse disorder including: salience, tolerance, preoccupation, impaired role performance, loss of control, and withdrawal symptoms. These features are consistent with those considered by the DSM-5. Validation of scales includes collecting normative data from a large and diverse sample of people and then developing a clinical cutoff. One might rightly ask if this usage pattern constitutes pathology or simply enthusiastic usage. To test this, researchers assess ``convergent validity.'' Specifically, how does the measured construct correlate with other outcomes we would expect it to predict. For example, we would predict that people with problematic social media usage would have increased risks of other mental health disorders (as is the case with other addictions). To that end, a recent metanalysis of 18 studies that assessed ``problematic social media usage'' with anxiety and depressive symptoms found a correlation of .348 (``medium'') and .273 (almost ``medium'') respectively. H. The Mechanism of Addiction Addiction (both behavioral and substance-based) is grounded in the brain's dopamine reward system. Exposure to a stimulus is processed in the Ventral Tegmental Area (VTA) of the mid-brain that releases dopamine. When that stimulus is ``favorable,'' the VTA signals the nucleus accumbens (the pleasure center of the brain) which in turn signals the Pre-Frontal Cortex (the executive center of the brain as discussed before) effectively saying ``I liked that'' so ``do it again'' or ``get more of it.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This is a generic pathway; it is activated for example when parents praise children for behaviors (e.g., saying ``thank you'') thereby increasing the probability (hopefully) that they will act that way more often. In pathological circumstances, given prolonged exposure to intensely pleasurable stimuli, the brain's natural reward pathways can be altered, making it increasingly difficult to experience pleasure from other activities. At a neurobiological level then, behavioral and substance-based addictions have a final common pathway. Functional magnetic resonance imaging (fMRI) studies have demonstrated that social media usage (and Facebook in particular) activates the accumbens. Indeed, Meta documents acknowledge that Facebook ``does activate the brain's reward system.'' Moreover, a three-year longitudinal study of 6th and 7th grade students found changes in functional activation of the brain based on reported habitual checking of social media sites at baseline. Specifically, habitual checkers' brains demonstrated differential activation of specific regions in anticipation of social cues compared to non-habitual checkers. This suggests that their brains may be being conditioned as a result of repeated activation (a feature of addiction). The causes of addiction are multifaceted, involving a combination of genetic, environmental, and psychological factors. Although not determinative, genetics may predispose individuals to addiction, while environmental factors such as peer pressure, stress, or trauma can trigger problematic substance use or behaviors. Mental health conditions like depression and anxiety are also linked to addiction, as individuals may turn to substances or behaviors as a form of self- medication. Early exposure to addictive substances, particularly during childhood, increases the likelihood of addiction later in life. I. Dangerous Features Social Media platforms embody numerous design features that promote addiction, problematic use, and attendant harms. This includes design features that exploit intermittent reward, social comparison metrics, and ``flow'' state. Each of these may discussed in turn, but they all work in tandem to keep users hooked and have their foundations in behavioral psychology, the ``father'' of which is BF Skinner. Skinner (1904-1990) was an American psychologist and the Edgar Pierce Professor of Psychology at Harvard University until 1974. He conducted foundational research related to how behaviors are reinforced through the ``Skinner'' box he invented. Briefly, rats were placed in a box that had a light, a loudspeaker, a response lever, a food dispenser and an electrified grid. The lights and the loudspeaker provided stimuli. These conditioned the rat to pay attention and press the lever when they were activated. Pressing the response lever would in turn dispense (or not dispense) food. The electrified grid could be used to ``punish'' failure to press the response lever when a stimulus had been delivered. Skinner found that rewarding reaction to the stimulus, by dispensing food if the rat pushed the lever when the light flashed, made the rat more attentive to the light. Conversely, ``punishing'' the rat for failing to respond to the stimulus by delivering a low-level shock via the grid also made it more attentive. Neither of these results are especially surprising (at least not now) to us. But the most interesting of Skinner's findings were that intermittent unpredictable rewards were more effective at inducing the rat to stay focused on the light than predictable ones. In other words, the rat became more attentive to a stimulus if not every reaction yielded a reward. This unexpected finding was long ago incorporated into such things as slot machines, where gamblers cannot predict which pull will result in payouts, and as they play, they are constantly hearing others around them win, reminding them that their next pull might be a jackpot. All three of these addictive design features are ``built in'' to social media sites and are reflected in the ``hook model'' studied by Meta's researchers: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This figure, which is taken from a book by Nir Eyal titled Hooked: How to Build Habit Forming Products, also appears in Snap's documents. The figure is annotated as follows: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] J. Likes, Comments, and Other Metrics Scientists have found that receiving ``likes'' on social media platforms are very similar to the ``rewards'' that researchers have associated with addiction research for decades. Likes can include the ``like'' button on Facebook and YouTube but can also include ``hearts'' on Instagram and TikTok. Similar metrics include the number of shares on a post, number of comments, and number of followers or friends for a user. While seemingly innocuous, users' quest for these publicly visible ``rewards'' has been linked to a number of secondary harms such as reduced sleep efficiency and duration due to ``routine check[ing]'' behaviors during the night. They have also been associated with reinforcing addictive behaviors in order to encourage users to spend more time on these platforms. They have even been linked to increased feelings of depression, anxiety, and negative social comparison since receiving ``fewer likes'' is viewed as a form of ``negative peer feedback.'' Comments are similar to likes in that they provide users with ``quantifiable (and qualitative) feedback'' about their experiences on the platform. For teens, this means that they are able to quantify the ``success'' of their posts with many teens reporting that they post ``self oriented images on social media with the goal of obtaining likes and other forms of feedback such as comments.'' Other studies have found that comments are often seen as a mechanism for ``gaining [social] status'' and reflect a degree of ``digital social approval.'' As a result, other studies have linked comments to negative social comparison since ``not receiving enough likes on one's pictures can negatively affect appearance esteem and prompt delet[ion] of a post.'' K. Algorithmic Recommendations At the heart of every social media platform's engagement strategy is their proprietary recommendation algorithms. These algorithms are optimized to maximize time on the platform rather than healthy interactions with a person's social network. They accomplish this in a variety of ways. One way they drive time spent on social media is by inferring the interests of the user (which may or may not be expressed by the user in any direct way) and feeding them an aggregation of posts that, while most likely to keep them engaged, may lead the user down problematic rabbit holes. Regarding TikTok, these problematic rabbit holes were also known as ``negative filter bubbles.'' A 2021 TikTok presentation entitled ``Negative Filter Bubble Project Sharing'' noted that ``Algorithms lead[] people into a feed filled with continuous similar negative videos that are hard to escape.'' What's more, another internal document notes that ``minor creators [fall] into higher density of negative comment filter bubbles.'' To its credit, TikTok did set as a 2023 safety goal to ``be more transparent to the public and trusted by teen users and parents,'' in addition to building ``industrial leading teens safety and well-being features.'' As discussed below, these aspirations of transparency and improved safety features that were noted approximately six years after TikTok launched in the United States were too little and too late for many children and teens. Another way recommendation algorithms drive time spent on social media is by utilizing the intermittent variable reward mechanism that Skinner discovered (e.g., likes, notifications, comments that appear unpredictably), which contribute to addiction to the platforms themselves. The addictive power of an engagement-optimized algorithm is recognized within the medical and academic literature. In the Handbook of Children and Screens, under the chapter regarding ``Problematic Internet Use,'' the authors' consensus was that ``Attention focused designs intended to generate, or possibly exploit, potentially addictive features (e.g., `likes') and conditioned responses (e.g., notifications) alongside powerful algorithm-based technologies may lead youth to stay online longer than either intended or recommended.'' L. Auto-Play Academics recognize that autoplay is a feature that is designed to prolong engagement at the cost of displacing important developmental opportunities for young children and is even associated with greater child behavioral difficulties. It has also been linked to difficulties controlling device use more broadly. The latter phenomenon has even been observed in social media companies' own research. For example, Meta's researchers learned that clinicians regarded autoplay as not having any ``beneficial'' role while ``detract[ing] from patients' ability to control amount of time spent'' using their platforms.'' M. Infinite Scroll Like autoplay, the Infinite or Endless Scroll feature has also been linked to prolonging engagement at the cost of displacing children's developmental opportunities. This feature is also linked to behaviors that create a user-sided ``time distortion'' that results in users spending more time on social media platforms than they originally intended. Some studies cited (and commissioned) by the social media companies have found that these effects are mitigated by ``active'' use of their platforms--e.g., generating content or posting content. However, this hypothesis has not been widely accepted, particularly since there is evidence that children are more likely to ``watch, play, or scroll through content created by others than they are to use their devices to produce their own content.'' N. Notifications Meta studied notifications and the ability for this feature to induce habitual or addictive behaviors. For example, one internal document supports the basic addiction principle that ``experiencing a reward (or reinforcement) can increase learning and motivation. This contributes to repeated, potentially habitual behaviors.'' The document included the following figure: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] It is not surprising then, that among the many metrics Meta tracks, the ``success rate'' of their notifications, defined as increasing daily usage, is a key one. Darius Kilstein, a Director in Data Science at Instagram, reports: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This is to say that Meta is monitoring (and presumably modulating) the intermittent reward mechanism to ensure that engagement is maximized. O. Beauty Filters Despite being a relatively new social media feature, Augmented Reality (``AR'') filters--commonly referred to as ``Beauty Filters''-- have been thoroughly studied due to the outsized negative impact they could have on users by exacerbating a socio-psychological phenomenon known as ``social comparison.'' Prior to the advent of these Beauty Filters, researchers had already identified that social media may exacerbate social comparison and lead to increased body dissatisfaction and/or disordered eating. However, with the introduction of Beauty Filters--many developed by social media companies--users are now exposed to ``manipulated'' photos that depict unrealistic (and in some cases impossible) body image Standards. P. Internal Company Documents Show They Knew What They Were Doing Notably, many of the features recognized in the literature as addictive and/or harmful were likewise identified in research conducted by Meta and presented in 2021. A Mixed Methods Clinicians study identified product features and pathways impacting mental health, including the following: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Of these features, several were identified as ``primarily negative,'' including Video/Photo filters, location sharing, automatically playing videos, and pop-up notifications. Internally, social media companies' documents recognize that they could increase engagement by changing the design of the social media. As early as 2016, Meta (then Facebook) was exploring ways to keep teens on its site and posting content: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] As Max Eulenstein, Vice President of Products, said in a Meta e- mail on January 26, 2021, ``No one wakes up thinking they want to maximize the number of times they open Instagram that day. But that's exactly what our product teams are trying to do.'' One mechanism that appears to be especially effective at engaging teens (and adults) is the use of ``reels'' or short videos that repeat. Meta adopted reels from TikTok after seeing how effective they were at promoting usage. Below are Instagram's metrics on the viewing of such reels by teens: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The data demonstrated that active daily U.S. teen Instagram users viewed on average almost 105 reels per day on the platform for a total of 22.5 minutes per day or an average of 20 seconds per reel per day. Briefly watching a short snippet algorithmically curated to one's interest is gratifying via the dopamine reward pathway. Further, their internal research identified the following ``Triggers for Problematic Use on Facebook'' that were later included as part of a ``Problematic Use Journey Map'' in 2020: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Each ``opportunity'' for a reward represents a potential algorithm tweak, and many--if not all--of these were eventually incorporated in some form into the platform. For example, a Facebook presentation contained following: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta's more recent documents depict a Venn diagram portraying one way to conceptualize the usage experience from a design perspective: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] ``Dark Pattern'' is a user experience term referring to interactions that are deceptive, or that trick you into doing something you didn't want to do. Again, this diagram acknowledges that Meta uses ``addictive'' design features and that some of them are set to ``default.'' Meta documents from Project ``Plato,'' intended to study and mitigate ``dark patterns,'' state that ``[s]ome UX [user experience] patterns rob users of their agency.'' Included among them are: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta's researchers also knew that being too critical of these design choices could lead to criticism from internal stakeholders. For example, Jennifer Guadagno noted that her wellbeing team's efforts to study Facebook addiction as part of Project Plato could ``get heavy pushback internally'' and added that she was not ``sure if it'll even be allowed to happen'' because ``if we now know all these things that are potentially bad and then we don't do anything to fix them'' it could lead to problems for the company externally. Arturo Bejar testified, in deposition at in open court, that Meta did not do enough to warn parents or curtail problematic use of their products and that the word addictive is `radioactive.' '' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Despite learning of these problems as early as 2018, Meta had not made meaningful changes to the platform nearly two years later. A Meta presentation from 2020 arrived at nearly identical conclusions as the Project Plato researchers, including the connection between design features and problematic usage. One slide from that 2020 presentation stated that research participants reported ``10+ triggers contributing to [problematic use] habits'' including: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta was not the only social media company to reach these conclusions about its platform features. For example, YouTube describes ``finding a video on YouTube search'' as a ``predictable reward'' compared to ``unpredictable rewards'' such as ``finding a new favorite song while in a mix,'' which the company described as ``disproportionately [more] delightful compared to predictable rewards because they're unexpected or exceed expectations.'' YouTube's researchers explaining the reason for this dichotomous design have stated: When thinking about habit building around YouTube, it is important that we reliably fulfill their goal pursuits (reliable reward) while also providing surprising nuggets of reward (variable reward). Ultimately, you're giving people more reasons to come back until they can't even remember why they did. When is the last time you had a goal in mind when you went to Facebook?'' Internal documents also reveal that Instagram is tracking usage/ engagement by teen users. An e-mail from Darius Kilstein, on February 5, 2022, states that Meta ``looked into the long-term decline of feed imp[ressions] for Teens,'' discovered that ``teens aren't relying on Feed for interest consumption as much as they used to,'' and hypothesized that this shift might explain ``why teens now consume fewer Reels than adults.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The reported ``headwind'' for the ``cohort'' points to Instagram's business need to redesign features to better engage younger children and keep up with their competitors (particularly TikTok), as Kilstein explained later in the same e-mail. The ``infinite feed'' invented at Facebook was exported and adapted to Instagram. In her deposition, Dr. Alison Lee stated: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The quote begins with the scenario whereby someone has been ``scrolling'' all day (itself an implicit acknowledgement of overuse) and exhausts their ``connected'' feeds at which point a new, end of feed algorithm is triggered. Aza Raskin is widely considered to have invented the infinite scroll. This feature allows a user to scroll indefinitely through their feed, receiving endless posts and intermittent variable rewards. Raskin testified that infinite scroll was like ``digital cocaine'' for the user. One of the technology leads for Instant Articles on Android at Facebook reported ``we only care about things like time spent, open links etc. That's what we optimized for. That's what we defined as success and failure.'' In 2022, Meta launched Project MYST, a prospective, longitudinal study to follow 4000 teens (13-17 years of age) using the highly respected National Opinion Research Center (NORC) at the University of Chicago for survey data and Meta logs of activity on their platform. Using validated measures for a variety of outcomes and actual usage patterns on their platforms, MYST appeared by all accounts to be a robust, well-designed longitudinal study--the kind that only industry could conduct based on their access to actual social media use data. Among other things, MYST intended to test the following negative hypotheses: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] While the initial plan for the study contemplated its results being released publicly, there has been no evidence that Meta ever publicly acknowledged that the study was conducted or released any results prior to it being produced in the course of litigation and revealed to the public as part of the K.G.M. civil trial in Los Angeles just this past month. Indeed, as the study moved forward, Meta employees seemed to become explicitly fearful of it being publicly released. Leaders directed researchers to refer to the study only by its full name, because they would ``really prefer not to have a cute acronym (MYST). That makes things sound more interesting when/if it leaks. Can we please keep our names as boring and uninteresting as possible.'' Perhaps more illustrative is the tension that is recognized in Meta's internal documents discussing MYST, which reflect that ``there is a fundamental tension between doing rigorous research to identify whether there are potential opportunities for change/intervention and advancing external/narrative goals.'' This tension with external goals is reflected in Meta's many decisions to not share publicly what their research or outside experts conveyed internally. External goals included the goal to ``[a]dvance the credibility of research that finds small or null correlations between social media well-being and/or well- being relative to other factors,'' and to ``deflate the conversations about research claiming causal connections between social media and mental health and well-being.'' Perhaps most critically, Meta's stated external goal was to ``[m]ake the issue bigger than social media. Research should be designed in a way that provides a holistic understanding of the relationship between youth well-being and technology and other factors (e.g., cultural and environmental factors that lead people to use technology and their use of technology, rather than focusing on Meta-specific product offerings.'' Thus, the impetus for their research appears to be advocacy or public relations rather than scientific discovery or public safety. Unfortunately for Meta, Project MYST revealed two key findings. First, the study found that children who had adverse events in their lives were more vulnerable to ``potential problematic use'' (i.e., a gentle euphemism for social media addiction). Second, researchers found that parental supervision made no difference whatsoever to the addictive impact of Meta's social media platforms. The study found there was no relationship between a parent setting time limits or useful safety tools and whether their child became addicted. It is little surprise, then, that Meta buried the data and prohibited the study's results from ever seeing the light of day--until litigation in California State by private plaintiffs thrust the evidence into open court. YouTube documents also provided evidence of problematic usage by virtue of design. In 2015, YouTube launched Autoplay on desktop (followed by a launch in 2016 to app), which automatically and continuously played the next video--without any need for user action. Autoplay set to on be default because it generates more watch time. Autoplay watch time doubles at night. As one YouTube employee explained, ``This was the single most impactful launch in YouTube history, with +8 percent desktop watch time, +4 percent overall watch time increase.'' In YouTube's own documents, making the platform ``addictive'' was a core design feature. This was consistent with the aspiration of the same presentation of ``Building the world's most powerful and delightful video consumption experience.'' YouTube internally acknowledged the potential negative effects of digital videos in a 2018 presentation entitled ``Literature Review: Effects of watching digital videos and viewer well-being,'' which called out problematic Internet use in particular: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The slide calls out problematic Internet use and states that it overlaps with ``addiction'' consistent with my belief that the entire continuum to the right of casual use (Figure 20) can be viewed as problematic. Furthermore, it highlights how ``autoplay,'' a key feature of YouTube, drives the ``just one more video effect.'' Slide 10 summarizes data from an internal survey of ``265 respondents'' that calls out the ``stickiness'' of the app and states that its interactivity and notifications ``causes users to feel that they must be aware of what is happening on the platform'' which ``keeps users on the platform longer.'' A 2018 YouTube presentation exploring the ``negative effects that video watching has on user's wellbeing'' included an express reference to addiction: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] TikTok's engagement algorithm, meanwhile, is frequently viewed as the most effective in the industry: ``more personalized,'' ``more accurate,'' and ``more diversified.'' The company emphasizes its effectiveness in its marketing presentations and touts that more than 53 percent of suggested videos are ``viewed.'' An internal memo citing the company's ``vulnerabilities'' reports: ``In a user survey of 2,300 users in February of 2020, when respondents were asked to give a score of 5 to indicate strength of agreement with the statement `I spend too much time on TikTok,' the average response was a 4.0.'' It goes on to say, ``some elements of persuasive design may be unique to TikTok; for instance, the fact that when you click the back button on your phone to leave the For You Feed/all, you get a prompt saying, `Tap again to exit,' which can be seen as increasing friction for users seeking to leave the app.'' And still later concedes, ``TikTok is particularly popular with younger users, who are seen as more vulnerable to online harms and the negative impacts of compulsive use.'' In a ``2021 TikTok for Good Business Plan and Vision'' document, the following is stated: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Notably, the plan acknowledges that ``addiction to technology is a ubiquitous problem.'' At the same time, comment [82] notes that ``addiction'' could be ``considered a very positive metric in our field.'' In other words, for all of its public posturing seeking to discredit or minimize the existence of compulsive or addictive use of screens, TikTok's internal documents acknowledge its existence and even allude to its ``value'' to the industry. TikTok's algorithms are widely considered the ``best in the industry'' because of its effectiveness at driving engagement. In pursuit of that end, TikTok made leaving the app more difficult than others. And for those times when people did succeed in leaving, TikTok refined its ``push'' approach to, among other things, get people back onto the company's platform. TikTok's ``push methodology,'' like every change they make to their platform, was subjected to A/B testing where rapid cycle experiments were performed comparing one version to another with users randomized to experience one or the other (i.e., no confounding) and with an eye to ensuring that core metrics were not adversely affected. Finally, Snap--the makers of Snapchat--innovated on the FOMO [``fear of missing out''] concept of maintaining ``streaks'' by developing the Snap Streaks feature, which it specifically designed to gamify its platform and drive usage and engagement. Streaks are built and maintained by two people ``snapping'' back and forth on consecutive days. The quality or content of the snap is irrelevant--it's all about ping-ponging back and forth consistently. Streaks were extraordinarily effective. Less than a year after they were launched, 22 percent of users had at least one streak, and the average highest streak value was 76.243. Snap Streaks were also particularly popular with younger users, as 47 percent of Streak users were under the age of 17.244. Of course, the true number was likely even higher, given how easy Snap--and every social media platform--has made it for children to lie about their age. Snap Streaks' importance to users was also tied to Snap Scores. While streak counts are private to the participants, every user of Snap has a numerical Snap Score displayed on their profile. The way that this score is generated is opaque, but broadly speaking it reflects how much they engage with the app and how much engagement they receive from others, including Streaks. For users, high Snap Scores are a way to ``show off their popularity.'' Snap's research found that ``users mainly used Streaks as a social status measure (to increase Snap Scores, have as many Streaks as possible) rather than a communication tool.'' This is consistent with Snap's conclusion that once users knew how to increase their score, it ``may make user care about (if not become addictive to) their [] score.'' However, Snap saw this as a positive feature of Snap Scores because it would ``encourage [users] to produce and consume more snaps.'' But at the same time that Snap was chasing the engagement increases offered by streaks, worries about the addictive effect of Streaks on children were growing. For example, focus groups results forwarded by the Senior Director of Corporate Communications and Public Affair to the Vice President of Global Public Policy found the following: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Another employee put it even more bluntly: ``we seem to have tapped into some mass psychosis where 17 million people must keep the streaks going.'' A 2017 study of Snap ``Power Users,'' commissioned by Snap to better understand features that drove usage, found that ``[f]or some, streaks have become a ``compulsive behavior'' and that they feel they are ``in too deep.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Similarly, an internal presentation on streaks tried to spin them as a positive tool for building friendships but nonetheless was forced to observe that Streaks can be ``really stressful'' and ``make[] it impossible to unplug for even a day.'' To better assess the harms of streaks, in 2018 Snap commissioned a survey of 790 users aged 13-24 (its core demographic): [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In Snap's telling, this study showed that that only a minimal number of users found streaks stressful. If these charts seem familiar, it is because Snap presented this same conclusion to the United States Senate, writing that the study showed ``the majority of our community did not indicate Streaks were a significant source of stress--but six percent did.'' However, both the methodology and Snap's spin on the results were deeply flawed. Given Snap's millions of users, 6 percent still represents hundreds of thousands if not millions of users. And to get to 6 percent, Snap disregarded the ``moderate'' stress responses entirely. Snap also, without justification, combined those who found streaks a little stressful with those who found they caused no stress at all. A more accurate assessment of stress levels might be that 21 percent of users experience at least moderate stress because of snaps. A more holistic statement of the survey's results would be that 54 percent of streak users found streaks at least a little stressful. The other problem with sampling ``users'' to assess experiential stress is what in epidemiological terms is called the ``survivor effect.'' Many people who found streaks intolerably stressful would have left the platform or opted not to participate in a study about how stressful they are. In other words, this approach leads to a biased sample. Much like asking people still in a baseball stadium when the game has gone into 13 innings if games take too long will underestimate the true proportion of fans who think they do, asking regular users of an app how stressful it is underestimates the unease it induces. The study also depended on self-reporting from young people. But as Jennifer Stout, Snap's Global Vice President for Public Policy, pointed out, ``Kids like a lot of dumb things and parents are always trying to regulate their activities for their own safety!'' Snap's spin on the 2018 study is further undermined by the fact that Snap employees, including CEO Evan Spiegel, continued to express concern that streaks were addictive and harmful for users: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] A 2023 document containing suggested answers for an employee Q&A with Speigel is clear that Streaks ``can cause confusion and/or anxiety. We know this from the millions of support tickets we get every week from people asking us to restore a streak they accidentally lost.'' Users themselves make clear that they found streaks addictive. As the Q&A answer explained, users' frantic commitment to streaks could be tracked in part by tickets seeking to restore lost streaks. But the answer actually understated how desperate users were to have their streaks restored. By 2021, Snap was receiving an average of 400,000 streak restore requests each day, making up 95 percent of the total volume of customer service contacts. In some cases, individual users reach out directly to Mr. Spiegel to express the harms of streaks. Snap also explicitly embraced the idea that social rewards were key to keeping users coming back for more, conducting research that ``confirmed the hypothesis that Posters are motivated to post because audience feedback is their ROI reward.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Indeed, Snap's research ``confirmed the causal relationship between receiving feedback (views and replies) and the poster propensity to post again.'' Following these conclusions, Snap's product team proposed tweaking ``Story reactions'' and replies to generate more engagement and posting. And just as feedback made users feel good, Snap's user research team found that not getting feedback was ``discouraging.'' Snap researchers found that 44 percent of Snapchat Story posters didn't receive any feedback on a given day leaving them more anxious and worse off than not having posted at all. Senior Snap employees were clear- eyed about the consequences of manipulating users' basic neurobiology and socioemotional responses. Responding to suggestions for ways to increase feedback rewards on Snapchat, Mr. Spiegel chimed in to say: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Similarly, when reviewing proposed changes to the way that streaks operate, Stephen Collins, a Director of Public Policy, expressed a similar sentiment, observing that ``[r]ewards are known to drive compulsive/addictive behavior among some vulnerable groups.'' Of course, Collins's solution to this was not to reduce the role that rewards-maximizing play in Snap's design but just avoid using the word. Snap also used other forms of notifications to drive engagement. In 2023, Snap decided to implement an ``always on badge'' to get more users to try Spotlight. This badge was a red dot that was always present on the spotlight icon, visually drawing users to it. It would only go away once a visitor clicked on Spotlight and would reactivate every 60 minutes, regardless of whether there was any new content or other change to actually notify a user of. This one design change raised the daily active use of Spotlight from 53 million to 71 million, the single biggest increase in daily active use ever seen in an A/B test. Snap was not only made aware of the risks of these features, but they were also told to remove them by outside social media experts (SME's) they consulted. In sum, nearly every social media company adapts addictive and harmful design elements from their competitors. Separate and apart from the third-party content underpinning the platforms, these design elements insidiously work to keep users--typically children and adolescents--glued to the screens for longer and longer periods of time. Yet social media companies routinely turn to Section 230 to shield themselves from liability for these design choices. Q. SMVLC Clients Impacted by Section 230 Courts' expansive interpretation of Section 230 beyond its statutory language or legislative intent has turned the Internet into a lawless no man's land. The best way to illustrate the real-life consequences of the expansive interpretation is to see how it impacts real people. I am therefore going to discuss 10 of the cases I have prosecuted where Section 230 has been invoked by social media defendants to claim absolute immunity from victims' claims and seek dismissal as a matter of law. a. A.S. v. Meta [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] A.S. was born on June 22, 2002, and grew up in Yaphank, New York. A.S. was a confident and happy child, who loved reading, writing, and helping people and animals. She dreamed about becoming a veterinarian. She was active in singing competitions and theater, enjoyed being in the spotlight, and looked for opportunities to shine. A.S. got her first internet-enabled tablet device for Christmas in 2012, when she was 10 years old. She did not have access to social media, however, as her parents regularly checked her device and took other steps to make sure that she was not accessing inappropriate content. They also had a house rule that all devices were to be kept in the hallway at night. A.S. did not initially mind these restrictions because she used her tablet primarily to play Webkinz and make Webkinz videos. Webkinz are stuffed animals that come with a code you can use to play Webkinz games electronically. But in 2013, everything changed. Meta purchased Instagram in 2012 for $1 billion dollars and, by the time A.S. started fifth grade, the Instagram social media product was rapidly increasing in popularity and rapidly evolving in terms of its design and product features. Meta made some of the most significant, and harmful, changes to its Instagram product from 2012 to 2016, including targeted advertising features, Explore, algorithmic changes promoting engagement over integrity, Reels, Stories, Live, the ``like'' button, view counters, easier access to multiple accounts, and direct messaging, to name a few. A.S. was exposed to and harmed by those changes without her parents' knowledge or consent. As A.S. approached fifth grade, it seemed like all her friends had Instagram accounts. She started getting teased for not having one, and her friends told her that she needed to open an account and could open one even if her parents said no. Instagram designed its product to provide access to as many people as possible, without regard to age or safety, and this was well known among children her age. It was also understood that Instagram wouldn't close your account for being under 13; you just had to say you were 13 when opening an account and could then put your real age in your bio. In fact, this is something many kids do even to this day. It was also understood that Instagram did not object to kids using more than one account, which made it easier to hide more personal content and the existence of secondary accounts from parents and family--which accounts Instagram and its young users refer to as FINSTAs (short for ``fake Instagram''). In 2013, A.S. was 11 years old and in the fifth grade. She opened her first Instagram account, without her parents' knowledge or consent. She could not access the account often since her parents monitored her tablet and did not allow social media. Nonetheless, on November 6, 2013, she wrote in her private journal, [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Instagram had every reason to know that A.S. was under the age of 13. A.S.'s first bio (which was often publicly viewable) read ``11 years young'' and when A.S. turned 12, she changed it to ``12 years young.'' She also regularly published her real age including, for example, in comments she left on other users' posts and/or pages: ``Hello, I'm 12 years old and love webkinz . . .'' ``I'm 12 and love them . . .'' ``I'm almost 13'' A.S.'s secret use of and developing addiction to Instagram coincided with a steady, but severe, decline in her mental health. As A.S. became more dependent on Meta's addictive-design social media product, she began to resent her parents for not allowing her to have an Instagram account and her inability to access that account at home. She also began engaging in harmful social comparisons, of the types identified by Meta in its studies. In November of 2013, she was starting to show signs of depression and her parents sought mental health treatment--they started her in counseling. They did not know that A.S. was using social media, and A.S. refused to continue seeing the counselor after a couple sessions. In May of 2014, Instagram provided A.S. with a solution to her access problem. Instagram featured user content advising A.S., and other children like A.S., on how to bypass parental controls. A.S. saw content from other users explaining how to obtain your parents' passcode as well as applications you would need to download to hide Instagram once you were able to get it onto your device. A.S. did just that. She used the information obtained from Instagram to get an application that enabled to her to make the Instagram application icon look like a calculator, which she then moved next to other utilities applications. A.S.'s parents continued with their periodic checks of her electronic devices but never realized that what appeared to be a calculator was really the Instagram social media product. As a result of Instagram's addictive design and A.S.'s addiction to the Instagram social media product, she lied to and tricked her parents. She felt guilty for her actions but also felt like Instagram was something she needed and that her parents were being unreasonable in denying her access. On May 23, 2014, A.S. opened her second Instagram account. She was still only 11 years old. Meta provided A.S. with access to this second account through a school-issued e-mail address, to which A.S. did not even have inbox access. That is, Meta has designed its product to not require identification or age verification and, also, Meta purposefully does not verify or check e-mail account authenticity, at least in part, so that it can claim plausible deniability as to the millions of young children using its application that are below the age of thirteen and/ or lack parental consent. In fact, when opening one of her later FINSTAs, A.S. got tired of creating fake accounts so she hit the keyboard randomly and used [email protected] as her account opening e-mail address, and Meta allowed her to open the account. Had Instagram not designed its product in this manner, A.S. would not have had access to multiple, secret accounts. Less than a month later, on June 9, 2014, A.S.'s parents caught her with her tablet in her bedroom. A.S. denied that she was doing anything wrong, and wrote about it in her journal, [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] At all times relevant, Meta knew that A.S. was under 13 and that she was opening multiple accounts under different usernames yet failed to restrict her access or notify A.S.'s parents, Kathleen and Jeffrey, of their daughter's account status. Meta also knew or should have known that A.S. began accessing her multiple Instagram accounts in the middle of the night and for several hours at time, which addictive and unhealthy behavior foreseeably compounded the harms to A.S. and her family and caused additional harms. Once again, Instagram did not notify A.S.'s parents or take any other step to restrict her access to its social media product. On the contrary, Instagram is aware of FINSTAs and considers the opening of multiple accounts by its teen users as a primary source of engagement and growth. For every minute A.S. spent on Instagram in the middle of the night, Meta earned more money. In fact, Meta is fully aware as to the times of day when users are on Instagram, how long they are on Instagram, and even that middle of the night usage is a significant indicator of addiction (which Meta prefers to call ``problematic use''). Meta simply doesn't care enough to do anything about it, and Meta's leadership has taken the position internally that engagement and retention are the top priorities, with user safety coming in distant third. Through her use of Instagram, A.S. also received explicit sexual communications and images from adult users and was messaged and solicited for sexual exploitive content and acts on numerous occasions by adult users of Instagram. These adult users are encouraged to use Instagram to sexually solicit and abuse minors due to Meta's refusal to verify identity and age for new users or implement feasible safeguards to protect minor users from receiving inappropriate sexual content. At all times relevant, Meta knew that some of its Instagram users would become addicted to Instagram (or, in Meta's words, would engage in ``problematic use''). Meta also knew that children and teenagers would be particularly susceptible, and Meta knew or should have known what an addiction like this would do those children and teenagers and their families. To name only a few examples, A.S. spent increasing amounts of time consuming the unhealthy content and product features Instagram served up to her. This meant waiting up and sneaking her device from the hallway after her parents went to sleep. It meant lying about what she was doing on her devices. It meant accessing Instagram even where her parents were in the same room, by pretending that she was playing Webkinz or similar, age-appropriate games--which required her to always be on high-alert to not get caught. The harm caused by this Meta-engineered addiction was made worse by the fact that A.S. was only 11 years old when she was first exposed to Meta's inherently harmful and addictive product. Once again, however, Meta didn't care. For years, Meta has been conducting studies meant to help increase usage and dependency by children under 13, like A.S. As a result of the social media addiction Meta fostered and encouraged, A.S. spent increasing and unhealthy amounts of time on social media, became sleep deprived, became anxious, and felt guilty about what she was doing to her family. Her mental health and relationship with her parents suffered greatly as a result. For example, any attempt by her parents to restrict device usage was met with extreme anger, including one time when A.S. punched a hole in the wall. And what had always been a close relationship with her mother rapidly evolved into one where A.S. began to see her mother as overprotective, irrational, and wrong. Instagram's addictive design and product features led to A.S. feeling like it was her and her devices against her parents, which only worsened her depression and anxiety. But also, and as Meta knows, the amount of time a user spends on Meta's social media products can correlate to worsening depression and other harms. In December 2013, A.S. got her first cell phone, but her parents got her a slide phone and not a smartphone. She had a lead part in the school play, which required her to stay late after school, so her parents wanted to make sure she could contact them if needed. Then in December 2014, A.S. begged her parents for a smartphone because her friends had one. They were able to get one cheap or for free through their phone carrier, so they agreed. But they took steps to install parental controls and did not allow her to take the phone in her room at night. Kathleen and Jeffrey also put parental controls on their home computer, again to make sure that A.S. was only using age-appropriate applications and features. A.S. used what she learned on Instagram to bypass those protections too and obtained even more access to Instagram because now she could use Instagram all day while at school on her cell phone. In April of 2015--just four months after getting her first smart phone--12-year-old A.S. drew an image of herself next to her phone with the words ``stupid ugly fat'' on its screen. She had a similar image on her computer, and these same types of words reflected in her thoughts, [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] A.S.' social media use coincided with a steady, but severe, decline in her mental health. She was addicted to Meta's product and spent increasing amounts of time on social media, specifically, perusing content recommended and/or made available to her by Meta, which increasingly included underweight models, unhealthy eating, and eating disorder content. From 2012 through 2015, Meta developed and implemented technologies and features designed to increase engagement (and its own profits), but which were harmful to users, including A.S. This included but is not limited to things like new advertising features, which allowed advertisers to target A.S. based on her age, location, gender, and other characteristics. Meta failed to exercise reasonable care to ensure that the advertisements were safe. Meta profited from its advertising practices and product features, while A.S. was exposed to and harmed by harmful advertising content. For example, a. A.S. was only 11 years old when she opened her first Instagram account and, as such, was particularly vulnerable and the impact of her developing addiction to Instagram was made worse as a result. b. Meta designed its Instagram product to default minors to public profiles, exposing 11-year-old A.S. to inappropriate sexual content and serving her up for access to complete strangers who Meta knew posed the risk of bullying and exploitation. c. Instagram's product design, including things like search and explore and photo editing and filtering features, and Instagram's emphasis on advertiser content, caused A.S. to question her appearance and worth, and made her increasingly anxious and depressed. d. Because Meta does not limit the time young users spend on Instagram, A.S. began stealing her tablet out of the hallway at night to access her Instagram account, resulting in severe sleep deprivation, which only made things worse. In 2016, Meta added several additional product features that Meta knew or should have known would be harmful to minor users of its Instagram product. For example, a. In February 2016, Instagram started enabling users to easily switch between multiple accounts. A.S. opened at least three more FINSTAs (secret Instagram accounts) in 2016 alone and Meta exposed A.S. to addictive design, harmful advertising, and other, Meta-backed content through those accounts too. b. In February 2016, Instagram added view counters to videos, increasing the social comparison harms already caused to A.S. by Instagram. c. In March 2016, Meta switched its feed from chronological to algorithm-driven ordering. A.S. no longer simply saw posts made by ``friends'' in the order they were made but, instead, Meta prioritized and escalated certain posts and content in A.S.'s Instagram feed. More specifically, it promoted and prioritized harmful content based on the determination that such content was more likely to keep A.S.'s attention, thereby increasing her use of its product and its resulting revenue. While Meta's prior product design was pushing constant social comparison and eating disorder content to A.S., the new design bombarded her with it and A.S.'s exposure to super thin models and eating disorder content increased exponentially. d. In August and November 2016, Instagram implemented Stories and Live product features, respectively, which again increased A.S.'s exposure to harmful content, as well as her social pressure to participate and engage with Instagram. e. In December 2016, Instagram implemented the ability to mark comments as ``liked,'' which was initially available only on the mobile application, which is how A.S. accessed Instagram most of the time by 2016--via her cell phone. The ``like'' feature was particularly harmful to young users, as Meta discovered, and this product feature resulted in increased feelings of depression, anxiety, and low self-worth. The more A.S. accessed and used the Instagram social media product, the worse her mental and physical health became, which eventually included a life-threatening eating disorder and suicidal ideation. A.S. had always been slender, but after Instagram's algorithm and design started pushing extreme weight loss content and bulimia purging instructions, and recommending pages featuring excessively thin models, A.S. became obsessed with her weight. From the outset, A.S.' Explore page was filled with nothing but overly thin models and thigh gaps. A.S. started looking into diets and healthy eating, which led to more harmful content. Eventually, Instagram's algorithms led her to the terms ``ana'' and ``pro-ana'' in hashtags Meta permitted in its Instagram user Stories. Pro-ana refers to the promotion of behaviors related to the eating disorder anorexia nervosa. As made clear through internal Meta documents, Instagram's provision of that material to A.S. was both harmful and the result of a design defect. Meta amplified and pushed A.S. toward harmful content through various recommendation mechanisms built into its Instagram product. For example, Meta sent A.S. recommendations to eating disorder and self- harm themed pages and groups. Meta also sent A.S. recommendations for ``friends'' who were, in fact, adult Instagram users either suffering from these mental health issues themselves or using the Instagram product to find and exploit young girls; and, likewise, Meta recommended A.S. to these same types of adult users, who then sought to connect with her. As made clear through internal Meta documents, these product features are harmful to a significant percentage of Instagram users, particularly teens and young women. Meta directed A.S. to other Instagram users such as ``skinny88,'' ``thinnfaerie'' and ``wannabe-s-kinny'' and sites such as ``#thinspo,'' ``#skinny arms,'' ``#skinny girl'' and ``#skinny body.'' A.S. did not seek out these individuals or sites but was rather directed to this content not based on her preference, but rather through Instagram's algorithmic feature explicitly designed to select the content that would trigger the greatest dopamine reaction in her adolescent brain. Content promoting unhealthy body images and anorexic behavior triggered a greater dopamine response in A.S.' frontal cortex than content that was healthy and benign, so Instagram affirmatively directed her to content it knew was harmful to maximize the time she spent on its platform and the corresponding advertising revenues Meta received. These are just two examples of the thousands of dangerous and harmful connections, recommendations, and recommended content Meta pushed to A.S. via its Instagram product: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta also harmed A.S. through specific product features like direct messaging and group chat. For example, Instagram's Direct Message and group chat features allowed adult users to message A.S. directly, and to create eating disorder and self-harm themed group chats, exposing A.S. (a child) to several of these adult users at one time. As made clear through internal Meta documents, these product features are harmful to a significant number of underage users. They also are not necessary to Instagram's operation and are features Meta can restrict and/or disable--but Meta makes calculated and economic-based decisions to keep them in place because Meta knows that restricting or disabling them would negatively impact engagement and revenue. At some point A.S.'s mother also discovered that she had an Instagram account, which resulted in more fighting and separation between A.S. and her parents. Ultimately, it was clear that Instagram would provide A.S. with means to access its product no matter what A.S.'s parents did and, by then, A.S. was already over the age of 12. There was nothing A.S.'s parents could do, other than to keep an eye on her account usage and to remind her that she still had to abide by the house rules relating to electronic devices. However, A.S. only disclosed the existence of one account--the account she considered to be her ``personal'' account--and none of her secret accounts. Instagram's provision of multiple accounts ensured that A.S. was able to continue hiding her most harmful Instagram activities from her parents. This is precisely what she did, and A.S.'s continued addiction and sneaking around led to even more arguments and even greater anxiety and depression. At one point, Kathleen started trying to catch her daughter so she could find out what was wrong and why her daughter's mental health was not improving. On one occasion she thought she saw A.S. sign out of her one Instagram account and onto another. But again, A.S. vehemently denied it: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta was providing A.S. with constant, harmful access to multiple Instagram accounts, and had designed and was operating Instagram in a manner that ensured that there was nothing A.S.'s parents could do to stop it. But also, Meta was concealing from the public what it knew about addictive design and its own efforts to addict young children and what it was learning about the harm its specific product features were causing to children and teens--the precise types of harms that were now happening to A.S. because of Instagram. A.S.'s parents had no way to even determine Instagram's role in A.S.'s ongoing battle with severe depression, anxiety, self-harm, and eating disorders and as a direct result, could not obtain the information that they needed to enlist professional help. By early 2018, A.S.'s parents were aware of the self-harm and had just started to become aware of the eating disorder. They were getting medical treatment for A.S. and rearranging their schedules to make sure they could help her in any way she needed. To help A.S. further, they purchased her a therapy dog, Draco, and got him trained--at considerable expense--so that he would alert A.S. and her family if A.S. engaged in harmful behavior. Draco has helped A.S. in so many ways and saved her life on more than one occasion. Then, on May 8, 2018, A.S.'s school contacted Kathleen and told her to come to school immediately. Kathleen was on her way to work, turned around, and called her mother and asked her to meet her at the school. Kathleen and her mother met with someone in the school guidance department, who told them that another student had sent some photos of messages A.S. sent and Instagram posts expressing a desire to commit self-harm and suicidal ideation. The guidance department employee showed Kathleen some of the documents but not all, and she still did not have the usernames on A.S.'s secondary Instagram accounts. Moreover, at that time, Kathleen's only priority was to keep her daughter alive, so she contacted Stony Brook Children's Hospital for a referral and took A.S. to Mather Hospital. Mather did not have an available bed but was the only viable option for the treatment A.S. required. As such, A.S. and her parents, Kathleen and Jeffrey, stayed in the psychiatric wing of Mather for five days, while waiting for an available bed. As documented in hospital records, A.S. was hospitalized for a total of ten days in connection with ``Anorexia Nervosa and associated habits of purging, as well as major depressive disorder and anxiety. Since outpatient treatment has begun, A.S. has continued to struggle. Over the past weeks these struggles have intensified including suicidal ideation. She was seen today after school alerted parents of various social media posts, including suicidality.'' (Referral letter from Stony Brook Children's Hospital, dated May 8, 2018). The following are just some examples of information that had been provided to the school. To the best of Kathleen's recollection, these specific examples were not provided in 2018 but were obtained from the school years later (in 2022), [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In May 2018, A.S.'s parents did not have access to A.S.'s ``personal'' Instagram account and did not know of the several other accounts to which Instagram provided her with access--or the harmful content contained in those accounts. A.S.'s parents were still struggling to understand what was happening and how they could help their daughter. After A.S.'s release from Mather Hospital, on May 18, 2018, she received home teaching for the remainder of her 10th grade year. A.S. had always done well in school and wanted to go to college and have a career, so her parents did everything they possibly could to prevent these events from derailing her education. They worked with teachers to get her assignments and tutoring, and enlisted family to help with home schooling. A.S. was able to finish 10th grade and keep her grades up only because of the incredible support and resources her family was able to provide. A.S. graduated from high school in June of 2020 and started St Joseph's University in September of 2020. Shortly after the start of the school year, however, A.S. relapsed and engaged in extensive outpatient treatment. A.S. is aware of the challenges ahead and works hard every day to not slip back into her eating disorder. A.S.'s addiction and resulting mental health disorders were the result of the unreasonably dangerous Instagram product Meta made accessible to her and that she used. Meta's products were not reasonably safe due to their defective design and inadequate warnings. To this day, Meta has actively concealed the fact and has ``sought to stonewall and block this information [information about the dangerousness of their products, especially to young users] from becoming public.'' \1\ Meta ``intentionally'' \2\ hid vital information in its possession from the public, the U.S. government, and foreign governments, including information relating to the safety of children and the role of its algorithms and other Instagram product features in causing addiction, depression, anxiety, eating disorders, and other harms. --------------------------------------------------------------------------- \1\ October 5, 2021, Senate Hearing Transcript, Mr. Chairman Blumenthal at 00:05:21. \2\ Id. at 00:29:25. --------------------------------------------------------------------------- Meta made false statements to the press and public, designed to cover up the inherent dangers of their products and, even when asked direct questions as to how those products ``impact the health and safety of our children, they choose to mislead and misdirect.'' \3\ --------------------------------------------------------------------------- \3\ October 5, 2021, Senate Hearing Transcript, Ms. Francis Haugen, at 00:32:20. --------------------------------------------------------------------------- The family did not discover, and could not have reasonably discovered, that A.S.'s addiction, depression, anxiety, eating disorders, and other harms were caused by Meta's unreasonably dangerous products until September or October of 2021. As a result of A.S.'s social media addiction and the harmful content and features Instagram relentlessly promoted and provided to her in its effort to increase engagement, she had to undergo professional counseling, in-patient programs, outpatient programs, and eating disorder programs, and she will likely require help in the form of a service dog for the rest of her life. She must stay in constant contact with her doctors, and fights to stay in recovery every day. A.S. will suffer permanent mental and emotional damages because of what Instagram has done. A.S.'s doctors have also advised that long term physical damage is likely. A.S. also requires the support of her parents and her service dog, Draco, and cannot live the independent and successful life she planned for herself. This situation is a direct and result of the social media addiction Instagram fostered and encouraged for its own financial gain and harms that resulted therefrom. b. Rodriguez v. Meta, et al. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] When she was nine years old, Selena Rodriguez was given a computer tablet to access the internet. Shortly thereafter, she downloaded Instagram, TikTok and Snapchat, and made multiple accounts without the knowledge or consent of her mother Tammy. Selena quickly became addicted to these apps and spent increasing amounts of time on social media. By mid-2021, she was on social media at all hours of the day and night and would get as little as two hours of sleep. She communicated with 2,500 users, all but a handful complete strangers and many of whom were adults. Tammy made multiple attempts reduce or limit Selena's use of social media which resulted in extreme anxiety and volant outbursts. When Tammy tried to confiscate Selena's devices, she would engage in self-harm or run away from home in order to access social media products on other devices. Selena's social media use coincided with a decline in her mental health and academic performance. As a result of her addictive use of these social media products, Selena developed numerous mental health conditions including an eating disorder, self-harm, and abusive behaviors toward her mother and sister. By the spring of 2021, she was experiencing severe sleep deprivation caused the constant 24-hour stream of notifications and alerts sent on these platforms. Through her use of Instagram, TikTok, and Snapchat, Selena was repeatedly messaged and solicited for sexual exploitive content and acts on numerous occasions by adult male users of Instagram and Snapchat, who are encouraged to use these platforms due to the social media companies' refusal to verify identity and age for new users. Prompted in part by Snapchat's ``disappearing'' message feature, Selena sent sexually explicit images using Snapchat. Based solely on the single Instagram account available, we have identified 25 specific instances in which Selena exchanged sexually explicit content with adult users of these social media platforms. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] On June 23, 2021, in response to a request from an adult Instagram user, Selena sent a video of her masturbating which my law firm reported to the National Center for Missing & Exploited Children. Tammy repeatedly sought mental health treatment for Selena, and one therapist told her that she had observed a patient as addicted to social media as Selena. On May 1, 2021, Selena wrote a friend on Instagram that ``I am crying i wanna die i feel everyone hates me i fight myself every day not to kms because i remember some people love me.'' On May 20, 2021, she wrote to another Instagram user, ``I hate my life baby I am sorry I almost left.'' Selena became suicidal and was hospitalized for emergency psychiatric care on May 26, 2021. Between May 26 and June 3, 2021, Selena was hospitalized for suicidal ideation. Her treaters observed excessive social media usage as a factor in her mental distress. Upon her return from the hospital, Selena told Unknown User No. 30 on June 9, 2021, ``Sorry just got home from Connecticut children's hospital for trying to kms.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] On the afternoon of July 21, 2021, Selena accessed her mother's supply of Wellbutrin, placed her phone on a table in her bedroom and turned on the video camera. Holding two Wellbutrin pills between her fingers, she looked straight in the camera, tilted her head back, and placed the pills in her mouth. She then took a gulp of soda out of a bottle, looked into the camera, made the ``Peace Out'' hand gesture and playfully stuck out her tongue. Minutes before she died, Selena posted her six-second video on Snapchat with audio from the NF song ``Paralyzed:'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] c. Nasca v. TikTok [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Chase Nasca was a 16-year-old junior at Bayport High School. He was an honor student and star athlete with no outward signs of mental distress. Chase opened a TikTok account and was involuntarily subjected to thousands of extreme and deadly videos advocating violence against others, self-harm, and suicide. These videos including material accompanied by audio chosen and/or recommended by TikTok. Although Chase was searching for uplifting and motivational content, TikTok, by design, directed thousands of suicidal videos into his ``For You'' page. The TikTok product directed Chase to videos co-created by TikTok that recommended suicide and were accompanied by dark, suicide-themed songs. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Some of the videos TikTok directed to Chase suggested that young people should end their lives by stepping in front of a moving train. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] On the evening of February 18, 2022, Chase rode his bicycle to railroad tracks near his home at Fairview and Railroad Avenue in the Town of Islip and had the following text exchange with a friend. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] At 6:35 PM, Chase was struck and killed by a Long Island Railroad locomotive while standing on the tracks at Fairview and Railroad Avenue. Following his death, Chase's parents accessed his cell phone and discovered hundreds of TikTok videos promoting suicide, depression, and self-harm. Chase's father, Dean Nasca, launched an online campaign entitled ``TikTok is Terrorism'' to publicize the pernicious role the Chinese Communist Party exercises over TikTok's parent company ByteDance and the financial impetus that allowed ByteDance to propagate its deadly business model. Dean also wrote this song to commemorate his son's memory and posted it to his website. See below and at https:// chasenasca.com/. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] d. DeSerio v. TikTok, et al. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Mason Edens was born on April 13, 2006, and lived in Centerton, Arkansas. Mason was an outgoing and happy young man. He was a good student and loved playing football. Mason began using the Snapchat and TikTok platforms around the age of 13. His mother Jennie actively monitored his social media use and limited his phone use. Following a breakup with his high school girlfriend on October 31, 2022, Mason sought out affirming material on TikTok. He specifically sought content on how to mend a broken heart. Instead, on his TikTok account in the ``For You'' feed, Mason had been shown video after video of depressing content. Before this point, his search history mainly included sports and outdoor content. At one point following the breakup, Mason looked up ``positive affirmations'' on TikTok and liked a video that featured sad music. Instead of showing Mason positive affirmations, TikTok's algorithm showed him depressing content that he did not search for. For example, Mason had liked a video with the audio, ``I wanna put a shotgun to my f---- mouth and blow my brains out''. Another described a plan to die by suicide. Another video he liked asked ``what are your plans for the future?'' and showed the video of a firearm discharging. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] TikTok targeted Mason with AI driven feed-based tools. It collected his private information, without his knowledge or consent, and in manners that far exceeded anything a reasonable user would anticipate or allow. It then used such personal data to target him with extreme and deadly subject matters, such as violence, self-harm, and suicide promotion. TikTok did not target Mason with this because he sought it out or even because TikTok thought he wanted to see it, but instead, because TikTok determined based on its collection and use of his personal and private data that he would not be able to look away. Likewise, Mason did not ask for these experiences or even want to see them. Because of the harmful dependency TikTok engineered and fostered, however, he was unable to look away. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Jennie noticed Mason's mental health deteriorating. He was often on his phone all night. She noticed an increase in the frequency and intensity of his anxiety. After weeks of him not getting enough sleep, she confiscated his phone on November 14, 2022. An argument broke out. Mason rushed up to his room and locked himself in his room. As his stepfather tried to get in to talk to him, Mason retrieved his brother's shotgun and shot himself in the head using the same method that TikTok had recommended and foisted on him. He was 16. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] e. Roberts v. Meta, et al. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Englyn Roberts was born on July 21, 2006, and grew up in New Iberia, Louisiana. She was the youngest of eight children. Englyn was a bubbly teenager. She loved spending time with family, eating at different restaurants, and traveling to different places. She enjoyed dancing, with hip hop and lyrical dance being her favorites, and had dreams of going to college and becoming a choreographer with her own dance studio someday. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] When Englyn was eleven years old, she was given her first cell phone and, shortly thereafter, she downloaded social media products. Her parents provided her with the cell phone so that she would have Internet access, could take photos, and could stay in touch with family and friends. Not for social media. Her parents required the access code to Englyn's phone and said that anytime Englyn changed that code she needed to provide them with the new one, which she did. Her father Toney regularly monitored Englyn's social media use, or so he thought. What Toney and Brandy did not know is that social media companies make sure that underage children can access their social media products, including by marketing to children, failing to verify age or identity (even when the children openly admitted to being underage on their public profile and/or in posts and comments), and permitting the opening of multiple accounts. It was understood among children that Meta wouldn't close your account for being under 13. You just had to say you were 13 when opening an account and could then tell people your real age. In fact, this is something many kids still do. It also was understood among children that Meta did not object to kids using more than one account, which made it easier to hide more personal content and the existence of secondary accounts from parents and family--often referred to as a SPAM account or, in Instagram's case, a FINSTA (short for ``fake Instagram''). Instagram, TikTok, and Snapchat were designed to frustrate and prevent parents like Brandy and Toney Roberts from exercising their rights and duties as parents to monitor and limit their child's use of their social media products. Tony and Brandy were not aware when Englyn's social media usage started, nor did they ultimately know which products she used or how many accounts she opened without their knowledge or consent. Englyn's social media use coincided with a gradual decline in her mental health. Englyn became addicted and spent increasing amounts of time communicating on social media, and engaged in the video, games, and reward features of the social media products. Shortly after Englyn got her phone, her mother realized that Englyn was staying up at night on her cell phone. Her parents thought she was watching videos on the Internet and chatting with friends, as they still did not know about Englyn's social media use. Englyn had always been a well-behaved child, so her parents trusted her to use the phone responsibly. When they realized she was staying up on the phone, however, they instituted a 10:00 p.m. rule for all devices: Englyn's phone had to be turned off by 10:00 p.m. each night. Englyn agreed and told her parents that she would make sure her phone was off by 10:00 p.m. Unfortunately, she was already addicted to social media by that point--which addiction her parents had no knowledge of--and which caused her to break the 10:00 p.m. rule to use social media. One night, Englyn's mother couldn't sleep and caught her awake with the cell phone on in her room. Brandy and Toney Roberts now believe that Englyn had been regularly staying awake and was suffering from sleep deprivation due to her use of social media, and Meta, Snap, and TikTok's failure to verify age and identity. These companies should not have been providing Englyn with access to their social media products at all as she was under the age of 13. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Regardless, from that point forward, Brandy Roberts took Englyn's phone downstairs with her each night when she went to bed. Englyn then started waiting until Brandy or both of her parents were asleep and would retrieve the phone so that she could use social media throughout the night. She was always careful to replace the phone before Brandy woke up in the morning and never got caught. Occasionally, Brandy Roberts would restrict Englyn's cell phone access, though usually not for more than a few hours or a day. Nonetheless, these efforts at exercising parental rights and authority caused severe reactions in Englyn. Because of Englyn's social media addiction, she would panic without her cell phone and beg to get it back. She would tell her parents that they could do anything, except take her cell phone away. She would promise to do her chores, bring up her grades, or anything else that might get her the cell phone back-- she said that she could not live without it. During these times when Englyn did not have social media access she would get anxious and depressed and would cry and sleep until she got her cell phone back. Once she had her phone back, everything seemed fine again. In 2019, Englyn was photographed at a family event using the Instagram social media product: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] What no one knew at that time was that, while Englyn appeared to be doing fine to her parents and family, in fact, she was being bombarded by Instagram, Snapchat, and TikTok with harmful images and videos (ranging from harmful social comparison content to violent and disturbing content glorifying self-harm and suicide). She was also receiving sexual and exploitative direct messages from strangers, which are allowed and enabled by all three of the social media products at issue, along with harmful connection and content recommendations--all of which would eventually kill her. These social media companies were not providing fun and safe photo sharing services, but rather, were dealing in incredibly addictive product features and pushing harmful algorithmically driven content, intended to keep Englyn hooked on their products by any means necessary. Meta, Snap, and TikTok's social media products also made harmful recommendations to and about Englyn Roberts, connecting her with strangers to increase their own engagement and thereby their own profits, which recommendations had nothing to do with any communication or informational aspects of social media. Meta, Snap, and TikTok's social media products also provided other users with unfettered access to Englyn through public profiles and features (in the case of Meta and TikTok), recommendation systems (in the case of Meta, Snap, and TikTok), and features that provided direct messaging access to Englyn regardless of her minor status, regardless of whether other users were on her ``friends'' list or equivalent, and regardless of duration, time of day, or frequency (Meta, Snap, and TikTok). In 2018 and 2019, Englyn began interacting with and then sharing content about depression, suicidal ideation, and self-harm via social media, unbeknownst to her parents and teachers. Meta, Snap, and TikTok also utilized algorithms and/or similar technologies to steer Englyn towards and otherwise promote and amplify harmful and unsolicited content. These companies are not only aware of their promotion and amplification of harmful content but knew or should have known that the harms caused by their marketing and amplification systems would be exacerbated by Englyn's sleep deprivation. These algorithms and similar technologies are designed to exploit these vulnerabilities caused by social media. In September 2019, Englyn and her friend began exchanging links to incredibly harmful and violent page and users they had found because of Instagram's algorithm and recommendation systems, which content Instagram hosted on its platform for years while young users, like Englyn, continued sharing and viewing the incredibly harmful and violent content with alarming frequency. The content can no longer be found on Instagram, as Meta finally enforced its own Terms of Service and deleted the account in December of 2021--but not before Meta's algorithms recommended this content to any number of children. In September of 2019, Englyn and her friend messaged each other with the harmful content recommended and/or promoted by Instagram. The posting person's username was gasstati0nparty though this user's content is posted under more than one username. In one of the videos exchanged on September 19, 2019, the woman hung herself with an extension cord attached to a door: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This is the kind of material Instagram's algorithm was recommending to 13-year-old girls in September of 2019, and it remained in Englyn's Instagram messages until December of 2021, more than a year after Englyn's death, when Instagram finally deleted this account. The sleep deprivation caused by addiction and amplification of harmful content to which Englyn was exposed through her addiction to social media products was a cause of her depression, anxiety, self- harm, and suicidal ideation. On June 13, 2020, Englyn wrote to Instagram username brayreggiontae_, ``right just think people be suicidal for no reason.'' brayreggiontae_ replied, ``And people still wouldn't understand.'' On July 15, 2020, Englyn wrote to Instagram username lexy.paull, ``but if i ever hurt myself tell my parents that it wasn't their fault it's my problems with finding love.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Around this same time, Englyn kept a list of mental health numbers (including a suicide hotline) in her ``private story'' feature on Instagram and was exchanging pictures showing self-harm with one or more of her ``friends'' on Instagram. Meta knew that Englyn was a minor and said and did nothing to alert her parents or teachers to what was transpiring on its platform and because of the harms caused by social media--harms Meta itself was already studying and knew about as being among the impacts and effects of use of the social media products at issue in this case. On August 11, 2020, Englyn typed an ``emotional description'' relating to her social media use into her cell phone's note pad application. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] On August 29, 2020, at 3:30 a.m., Englyn took an extension cord and attached it to her door--just like she has seen in the Instagram video viewed countless times by herself and other young children--and attempted to choke herself--just like she had seen in the Instagram video viewed countless times by herself and other young children. Englyn texted her boyfriend (with corrections and as taken from police report): I just want to die. I can't satisfy anyone. In the closet choking myself, wonder if I'm good enough to stay on Earth, because it's never enough, for anybody, nobody calls me, to check in on me, nobody, not even you, I have no reason but to die, call or I'm stepping off this stool, and I'll be done just like you wanted, never mind, see, take too long, dropping phone, so I won't get what you're saying, bye love you. She then used her phone and took a video of herself crying, in which the extension cord can be seen. Brandy and Toney Roberts received a text message from Englyn's boyfriend's mother at 3:30 a.m., asking them to check on Englyn. They found their daughter hanging from the extension cord moments later. Toney picked her up and laid her on the floor, and Brandy began doing CPR, which she continued until the ambulance arrived. The medics were able to get a pulse, and Englyn was rushed to the hospital, where she stayed on life support for nine days. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Her parents stayed by her side every possible moment, singing to her, playing music, and holding her hand. On September 7, 2020, they made the difficult decision to take Englyn off life support and continued to stay by her side as she left them. Englyn's death devastated her entire family and her friends. She was a bright light and loved by everyone who knew her, and Toney and Brandy had no choice but to assume that her death was caused by some sort of mental illness no one knew about and with no outward symptoms. They became active and vocal on issues of teen mental health awareness and made a commitment to help other families and children if they could. They also tried to reach out to Englyn's friends and other children on social media through postings and messages of positivity and hope. Then, one week after the anniversary of Englyn's death--the week of September 13, 2021--Toney Roberts was watching the news and learned about the Facebook Whistleblower, Francis Haugen. He began looking for and listening to news on this topic, learning about what Instagram's leadership knew about the harms its products are causing to teen users. To name only one example, Meta recognized that a significant percent of its users are addicted to its products, that children open multiple, secret accounts (referred to by Meta as a ``unique value proposition''), and that use of Instagram increases thoughts of what Meta calls ``SSI'' (Suicide and Self-Injury) in a percentage of teen girls who use it. Englyn was nothing more than a statistic to these companies, and Meta, TikTok, and Snap chose profits and growth over the health and well-being of unsuspecting teen users like Englyn. That was when Toney Roberts charged Englyn's old cell phone to see whether the things they were describing on TV were what happened to his daughter. He was able to access some of her social media accounts from her phone and began scouring those with the information now in his possession. What became clear in September of 2021 is that Englyn's death was the result of injury caused by her addictive use of Instagram, Snapchat, and TikTok. What her parents also did not know at that time was that, as a result of her use of Instagram, Snapchat, and TikTok, specifically due to the intentionally addictive nature of these products as well as the harmful content they were promoting and amplifying to Englyn via various algorithmically drive product features and harmful social comparison features, Englyn Roberts had begun to suffer from depression, anxiety, self-harm, and suicidal ideation. Toney and Brandy had no way of knowing what social media was doing to their child. In sharp contrast, Meta, Snap, and TikTok knew or should have known that they were causing this harm to Englyn, including based on her age, usage information, and usage patterns that were repeatedly exposing her via unsolicited methods, such as push notifications and algorithmically driven recommendation systems. Without Meta, Snap, and TikTok's failure to conduct a reasonable age verification, Englyn Roberts would not have been exposed to the harmful features and design of their respective social media products. Without the algorithmic discrimination contained in all three of the social media products at issue, including recommendation systems as well as content promotion and display systems, Englyn Roberts would not have been targeted and overwhelmed by disproportionately violent, sexual, and other harmful content. Without the endless feed and explore features characteristic of all three of the social media products at issue, Englyn Roberts would not have experienced the harmful dependencies that these features were designed to promote. f. V.V. v. Snap, et al. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] V.V. opened her first Instagram account and at age 10 and her first Snapchat account at age 12, both without her parents' knowledge or consent. V.V.'s parents believed she was using her iPad for games, kid- appropriate videos, and educational purposes. They would never have allowed Instagram in and Snapchat their home had they known she was using those apps. By the time V.V.'s parents learned she was using Snap and Instagram, she had learned how to open multiple accounts to thwart their exercise of parental authority. V.V. used fake e-mail accounts to open new Instagram and Snapchat accounts and provided inconsistent birthdate information across related accounts. Meta and Snap do not verify phone numbers, e-mails, or other information users provide, making it easy for underage users such as V.V. to open accounts and for predatory adults to use false identities to exploit vulnerable minors with anonymity and impunity. Meta and Snap knew or should have known that V.V. was underage and did not have parental consent to use their products; yet continued to provide her with new accounts profiting handsomely as a result. V.V.'s parents tried to stop her from using Snapchat and Instagram and, when that failed, opened their own accounts to monitor and protect their child. They understood that Snapchat was a silly photo app for users to interact with people they already knew; and Instagram was a product kids used to keep in touch with friends and family. V.V.'s parents had no idea that Instagram and Snapchat were connecting their 12-year-old daughter with strangers and sexual predators. Shortly after opening her Instagram and Snapchat accounts, V.V. began having trouble sleeping and, over time, experienced severe sleep deprivation and related harms such as anxiety, depression, moodiness, and inability to regulate her emotions. While V.V.'s parents did not realize she was using Meta and Snap's products during sleeping hours, Meta and Snap track the hourly usage of their products on a per-user basis. Meta and Snap specifically designed their products to hook young users such as V.V., utilizing a variety of extended use designs--which designs were known to Meta and Snap, but not ordinary consumers. Meta and Snap inundated V.V. with push notifications, which had the intended effect of pulling her back onto their social media apps at the expense of her mental health and wellbeing. V.V. became dependent on Meta and Snap's features, including gamification features and rewards, and often found herself losing track of time while using Instagram and Snapchat due to their pull to refresh and endless scroll features. Shortly after the opening of these accounts, Meta and Snap also began targeting V.V. with harmful content and advertising. Meta made available to her certain product features that were known to cause social comparison harms in teen girls. As direct and result of Meta and Snap's actions, V.V. experienced increased anxiety and depression, as well as devastating harms to her sense of self and self-esteem. These harms made her even more vulnerable to outside influences than she otherwise would have been by reason of being a child. V.V. also began to suffer harms resulting from Meta's ``like'' product and the volume of unsolicited health and beauty content Instagram sends to young girls. V.V.'s Explore page often was filled with images beautiful and skinny models and advertisements for beauty products and weight loss tips--not just a few of these, but an unlimited feed filled with them, to the point where it was hard to see anything else. Like millions of teens who use Instagram, V.V. worried and judged herself based on the ``Like'' button. She would spend hours trying to create the perfect photo and once posted, her self-image would rise or fall on how many likes it received. V.V. judged herself and believed others were judging her by these Instagram-set standards, and her self-esteem declined while her vulnerability increased as a result. Snapchat and Instagram target underage girls for unsolicited advances by predatory adults. When V.V. first started using Snapchat, she did not have her own account and used her grandmother and aunt's accounts. When V.V.'s Snapchat profile was of an adult woman, she was not approached by strangers, no one attempted to exploit or abuse her, and no one sent her explicit Snaps and messages. However, once V.V. opened her own Snapchat account and identified herself as a teenage girl everything changed. She immediately began receiving sexually explicit images and solicitations from predatory strangers. Through Meta's public profile features, V.V.'s profile was made viewable and available to complete strangers (including adults) also using Meta's products. User recommendation systems--known as ``People You May Know (PYMK)'' on Facebook, ``Suggestions for You'' on Instagram, and ``Quick Add'' on Snapchat--also immediately began sending V.V.'s profile to adult male users, and encouraged those complete strangers reach out and connect with V.V. Snap has gamified its platform to incentivize and encourage children to accept friend recommendations from predatory adults. For example, a user's Snap score--a matter of social prestige and pressure among minors--goes up with each new connection accepted and the Snapchat product previously allowed and encouraged minors to accept all Quick Add requests via a single click of a button. Snap and Meta have designed their products to not only make these connections, but to ensure that their youngest users accept them. Yet V.V.'s parents had no reason to suspect that Meta and Snap were trafficking their daughter through complex and proprietary algorithms, profile settings, and other features. As soon as she opened her Instagram account, adult male strangers began finding V.V. through Meta's ``Suggestions for You'' feature and then would try to add her and/or would send her direct messages to connect. When V.V. was 11 years old, Meta directed and connected her to one such predator--a user whose real identity is still unknown. This Instagram predator seemed harmless at first, but his messages quickly turned explicit and he coerced V.V. into sending him explicit photos of herself. This was something that she couldn't tell her parents about, especially because they did not know about her use of Instagram. When V.V. tried to suspend contact with this Instagram predator, he threatened to upload the photos to an Internet site unless she continued to engage with him and sent him more photos. V.V. was terrified of her parents finding out, and of her photos being published online, so she sent him more explicit photos. V.V. felt trapped, and in December of 2018, she tried to commit suicide in the hopes of finding some escape. V.V.'s parents did not know the reason behind her suicide attempt at the time because V.V. was still afraid to tell them the truth. When she finally did, they attempted to report the exploitation to the police, who informed them that Meta's Instagram product was designed in such a way that they could not identify who was behind the account that was used to abuse her, and a case was never opened. Through this incident, V.V.'s parents learned she was using the Instagram social media product and took away her electronic devices to prevent such use. They noticed that when V.V. did not have access to electronic devices, she began sleeping better, would engage with her family, and participate, and was less anxious. Unfortunately, however, they were unable to prevent V.V. from using electronic devices indefinitely, and as soon as she was able to gain access again, she resumed using Meta and Snap's social media products. After V.V. opened her own Snapchat account (instead of using the one belonging to her grandmother and aunt), male strangers started sending her photos of their genitals. The way the Snap product is designed, V.V. could not discern what the photo contained until opening it. When this first started happening, V.V. reported these incidents to Snap. She forwarded Snap the explicit photos she was being inundated with, but Snap did nothing in response. On the contrary, often the same male user would continue sending V.V. explicit photos which she would report, but again, Snap never responded. On July 15, 2019, when V.V. was still 12 years old, Snap connected her to Reginald Sharp. V.V. did not know Sharp in real life and Snap directed him to her via its Quick Add and/or Public Profile product features. Sharp, a registered sex offender, went by the username JASONMORGAN5660. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] After Snap connected V.V. to Sharp, he quickly began engaging V.V. in sexually explicit communications and promised her money in exchange for sending him explicit photographs or meeting in person. After V.V. sent Sharp explicit photographs, he threatened to post those on Snapchat unless she had sex with him. On July 23, 2019, Sharp coerced V.V. into sneaking out of her home in the middle of the night to meet him and raped her. V.V. was 13 at the time of her rape. V.V.'s mother knew something was wrong immediately the next morning, found out what had happened, and reported it to the police who investigated. Detectives identified the Snapchat predator as Reginald Sharp and served a subpoena on Snapchat. Nevertheless, even though Sharp is a known sex offender who raped a 13-year-old child, as late as 2023 Sharp's Snapchat account was still active, as shown below. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] V.V. has suffered serious mental, emotional, and physical harms and been hospitalized as result of Meta and Snap's marketing, design, distribution, advertising, and programming decisions. She has isolated herself socially, suffered from anxiety and depression, thoughts of self-harm and suicide and feels tremendous pain and guilt for the turmoil her social media dependency has inflicted on her family. g. Arroyo v. TikTok [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] TikTok relies heavily on the amount of time users spend watching videos to steer users toward more videos and continuous scrolling. This process sometimes steers younger users toward material promoting self- harm and/or suicide. In December 2021, an anonymous TikTok employee publicly shared an unauthorized, internal document because they were disturbed by TikTok's push toward ``sad'' content that could induce self-harm. The internal communication called ``TikTok Algo 101,'' revealed that TikTok's goal was to keep users on the app as long as possible. The document indicated that one solution to getting a bored user to continue watching, is through ``forced recommendation in users' ``For You'' feeds.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] A deadly ``blackout challenge'' on TikTok was linked to the deaths of at least 20 children between January 2021 and November 2022. Children around the world received the blackout challenge in their ``For You'' feeds, encouraging children to choke themselves with household items until they become unconscious and film the adrenaline rush once they regain consciousness. TikTok knew children were dying from this challenge that was delivered to children via its algorithms and reaching those too young to understand the dangerous risks involved with participation. At least 15 of the 20 children who died while filming the challenge were aged 12 or younger; one of them was nine- year-old Arriani Arroyo. Arriani Arroyo lived with her parents, Heriberto and Christal Arroyo, and her brother, E.A., who was three years younger. Arriani enjoyed playing basketball, kickball, and riding her bicycle. She loved to dance and was enrolled in a ballet class at school. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Arriani began using TikTok at around age seven. She used TikTok multiple times a day, including watching videos of other people dancing and singing and posting videos of herself dancing and singing. Arriani gradually became obsessive about posting dance videos on TikTok and became addicted. She began receiving from TikTok challenges on her ``For You'' page. She sometimes discussed these challenges with her parents and because all the challenges they discussed involved eating and drinking, they seemed harmless to Arriani's parents. They understood that TikTok was a family-oriented social media product, which was marketed to and safe for children to use. On February 26, 2021, while her mom was attending a church event and her dad was working on a project in the basement, Arriani played with then five-year-old brother, E.A., upstairs in Arriani's room. E.A. went downstairs and told his father that Arriani was not moving. Heriberto rushed upstairs and found Arriani hanging from the family dog's leash, which she had affixed to the door to her room. Heriberto Arroyo called 9-1-1 and Arriani was rushed to Children's Hospital where physicians placed her on a ventilator and were able to restore her pulse. However, testing revealed that Arriani had permanent, irreversible, and complete loss of brain function and life support was withdrawn. h. Neville v. Snap [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Alexander Neville was born on May 4, 2006, and lived in Aliso Viejo, California. He loved skateboarding, videogames, and playing with his little sister, E.N. He had a keen interest in World War II and the Civil War and often shared with his family and friends his dream of becoming a historian when he grew up. Alexander got his first cell phone in the 5th grade. He started attending the YMCA for afterschool care and his parents wanted him to have a way to reach them if needed. They talked about responsible cell phone use with Alexander and told him that it was okay for him to play age-appropriate games. His parents also required Alexander to provide both his device pin and passwords to them for any applications he put on the device, so they could check in on him. Alexander's mother, Amy Neville and her husband Aaron Neville believed that Snapchat was a social media product made for kids. They understood that it was made for 13-year-old kids, and that it was a fun and silly product where you could take goofy photos and share those with friends. They knew that most 13-year-old kids already had a Snapchat account; in fact, the director of the YMCA said that she had opened a Snapchat account so that she could keep tabs on her YMCA kids. Snapchat was advertised as and appeared by all accounts to be harmless and age appropriate. In retrospect, Amy and Aaron now believe that Alexander began using Snapchat sometime just prior to starting the 8th grade, if not earlier. This was around this time when they first observed Alexander's trouble sleeping. For example, Amy would find him awake in the middle of the night, in his room and on his phone. Throughout 8th grade, Alexander struggled more than usual. He was often tired, and his parents resorted to the threat of turning off wi-fi to get him to sleep. Alexander's use of Snapchat coincided with a steady decline in his mental health. As a result of Snap's products and features, i.e. push notifications, user recommendations, interface and operational extended use designs, rewards and gamification features, etc.--Alexander began suffering from severe mental health harms, including, but not limited to, social media compulsion, sleep deprivation, increased anxiety, and depression. Snap also began directing and recommending drug advertisements to Alexander and connecting him to Snapchat drug dealers via its recommendations and mapping and location features, as it does to millions of underage Snapchat users. He received multiple Quick Add requests from other Snapchat users he did not know in real life and, because of Snap's gamification and other features that encourage kids to accept those requests, he accepted them. Among the strangers to whom Snap connected Alexander were nearby Snapchat dealers--persons Alexander did not know in real life and would not have met but for Snap's product design, programming, distribution, and operational decisions. Snapchat designed and programmed its user connection systems to increase engagement at any costs which, in the case of minor users, included affirmatively identifying and directing predators and drug dealers to those minor users. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] On June 21, 2020, just a few days before Alexander died, Amy Neville asked him what was wrong, as she could see that something was off. Alexander told his mother that somebody on Snapchat had sold him an Oxycodone pill which he'd taken. He said that he was curious about it but now, he said, he was scared because he already wanted more, and that he needed help. Amy called a drug treatment center on the morning of June 22 to get help for her son. On the morning of June 23, 2020, Amy went to Alexander's room to wake him up for an orthodontist appointment. She opened the bedroom door and found Alexander's body laying lifeless on his bedroom floor. Alexander died of fentanyl poisoning at the age of 14. Amy was made aware that the pill that had been sold to Alexander was 100 percent fentanyl, and that he obtained the pill through Snapchat dealer AJ Smokxy. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The authorities subpoenaed Snap for documents, and it took Snap nine months and multiple subpoenas to finally answer. Once Snap finally responded, there was enough information to confirm that Alexander met with the Snapchat sealer AJ Smokxy. Even after being served the subpoena, Snap continued to let AJ Smokxy sell drugs on Snap. They did not disable or block the Snapchat dealer who sold Alexander counterfeit Oxycodone, despite knowledge that AJ Smokxy was selling deadly fentanyl pills to minors via and because of its Snapchat product. In August of 2020--two months after Alexander's death--Snap was still facilitating the drug deals for AJ Smokxy. AJ Smokxy was also publicly denying his role in Alexander's death while publicly admitting that he did deal ``jars'' and ``bud.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Snap allowed AJ Smokxy to keep marketing and distributing his drugs through its platform--and Snap continued to profit from its casual partnership with him. Snap caused the deaths of at least two more individuals after Alexander because of drugs distributed by AJ Smokxy on Snapchat. Snap knew AJ Smokxy was under investigation for drug- related homicide but continued to let him market and distribute his drugs through their product, failing to delete his account even after they were on notice that he was killing kids. To name only one example, AJ Smokxy went on to use the Snap drug network to supply counterfeit pills to the Snap dealer, Arnoldo_8286, who found and contacted Daniel Elijah Figueroa (``Elijah'') via Snap and sold him a counterfeit pill. Elijah died three months after Alexander, in September 2020. AJ Smokxy's account remained active for roughly a year after Alexander's death and when Snap would have received subpoenas relating to that death. Snap knew that AJ Smokxy was using its Snapchat product to sell drugs to young Snapchat users. Snap knew or should have known that AJ Smokxy was using its Snapchat product because of its unique product features like disappearing messages and My Eyes Only. Snap knew that it was benefitting financially from his Snapchat activities and from its refusal to deactivate his account. Snap made the choice to take no action in connection with these illegal activities, and to disregard its promises as per its terms of service. It did not de-activate the Snapchat Dealer account or warn users or their parents of known illegal and potentially lethal harms arising from such use and other Snapchat users were harmed and/or died a result of those decisions. i. N.M. v. Meta et al. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] N.M. was born into a devout Irish-Catholic and tight-knit family that never misses Sunday mass or lunch with her grandparents afterward. N.M. was a happy and fiercely intelligent child. She enjoyed and did well in school and sports and dreamt of attending college on a field hockey scholarship and studying to become a pediatrician. She has always lived, except while attending college, with her parents, E.M. and P.M., and three siblings. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] N.M. got her first cell phone in 8th grade, after saving up and purchasing it herself. She was active in sports, a straight-A student, and a responsible individual who gave her parents no reason to think that she could not handle the responsibility of a cell phone. Her parents discussed with her the importance of safety, of not sharing personal information with strangers, and of being careful with what she posted with the knowledge that once she put something out there, she could never take it back. One of the first things N.M. did after purchasing her phone was to open an Instagram account. It seemed like all of her friends already had Instagram, and she opened the account to keep in touch and communicate with her friends and family members. Around late 2019, N.M. opened a TikTok account. N.M.'s use of Instagram and TikTok coincided with a steady and severe decline in her mental and physical health. N.M. was not looking for and did not want fitness, weight loss, exercise, or any similar content or advertisements from Instagram and TikTok. She was athletic, engaged in several sports, and had a good sense of self and self- esteem--then Instagram and TikTok happened. Instagram sent push notifications 24 hours a day designed to lure N.M. back to the app every time she tried to log off. Within a matter of months, she found herself regularly checking her Instagram at night, at first in response to push notifications Instagram sent her and, eventually, as a matter of compulsion. She intended to spend only a few minutes on the product but would get lost in Instagram's endless scroll and other extended use product features, lose track of time, and ultimately spend hours passively viewing content on the app, causing a lack of sleep, anxiety, and depression. N.M.'s posting experience quickly changed, going from something she enjoyed to something she obsessed over and, at times, dreaded. She worried relentlessly over the number of likes, comments, and shares her posts received, would check her account multiple times after posting, and would remove posts if she felt they did not garner enough responses. Instagram created a perfect storm of social comparison, which left N.M. feeling anxious, vulnerable, and often unsure or insecure about herself. These harms were caused specifically by Meta's product designs, harms in which Meta knew about and opted to inflict on a significant number of Instagram users (specifically, teen girls) to ensure increased engagement. Over the course of those first few months, Instagram began selecting and connecting N.M. with harmful users and content. This included random strangers who Instagram would recommend to N.M. via its ``Suggestions For You'' user recommendation algorithm as well as random strangers who were somehow about to find N.M.'s profile and direct message her even though she was a minor with a profile set to private. These product features were either defective or inherently dangerous in that they were serving up a minor to predatory, adult users. Instagram likewise began selecting and connecting N.M. to massive amounts of advertising that made her feel like she wasn't pretty enough or skinny enough. At first, this included makeup advertisements, then it became advertisements for workout products, exercise products, and fat burning and/or low-calorie food products. These were advertisements Meta either directly approved or assisted in the creation of, then purposefully targeted at N.M. due, at least in part, to her age, gender, and other perceived vulnerabilities, and from which Meta profited each time. N.M. did not search for these product advertisements and did not understand why Instagram was targeting her with them; however, she then began asking her parents to buy these products. Instagram then started targeting N.M. with large amounts of Health & Beauty connections and content, including influencer and other third- party videos focused solely on health and beauty, and similar subject matters. These were subject matters N.M. did not search for and did not want, but that Instagram began pushing to her Explore and other app surfaces in excessive amounts each time she opened Instagram. As with the advertisements Instagram directed to N.M. via its product design and programming decisions, Instagram began by targeting her with bathing suit photos and videos, then cooking recipes and ideas, and over time, Instagram started filling her feed surfaces with overly skinny models, more make-up products, and diet products and tips. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Instagram targeted N.M. via its product design and programming based on her age and gender, and the decisions Instagram made for N.M. caused incredible harm, including but not limited to harmful social comparisons, depression, anxiety, and, eventually, disordered eating as N.M. began buying the products and following the diets Meta told her she should buy and follow. Like Instagram, TikTok also sent N.M. push notifications 24 hours a day, designed to and successfully did persuade her log back on to the TikTok app every time she tried to log off. N.M. found herself opening her TikTok app regularly and at all hours due to extended use designs utilized by TikTok, including not limited to auto-play, endless scroll, and time manipulation techniques. What N.M. meant to be a few minutes on TikTok almost always turned into hours, which developing dependency led to increased sleep deprivation, anxiety, and depression. N.M. began watching TikTok for its cute dance videos, but within a matter of months--if not quicker--the TikTok algorithm product also began to turn deadly. Like with Instagram, TikTok started by flooding N.M.'s For You Page with advertisements for makeup and health and beauty brands. Then it began sending N.M. skinny influencer content, encouraging and pushing her to try new workout routines--which, ultimately, she did; then TikTok targeted N.M. with extreme content like ``What I Eat In a Day'' and dangerous workout routines. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] TikTok also materially contributed and participated in the creation of most if not all of the influencer (referred to by TikTok as ``creator'') content it then directed to N.M., including by encouraging and paying those creators to post more, telling them where to post, when to post, and how to post, providing them with tutorials and coaching them on how to get more views and make their posts more interesting, suggesting songs to match their themes, and other tools, communications, and actions in which TikTok routinely engages specifically with those users its refers to as Creators in order to make more money via engagement. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] When N.M. opened her TikTok account, she provided her actual birthdate. TikTok knew or should have known at all times that it was directing these advertisements, creator content, and other harmful subject matters to a minor. Meta also knew or should have known that N.M. was a minor, including based on information it had in its possession, its age estimating technologies, and N.M.'s stated birth year, which, while not accurate, still identified her as a minor when Instagram began targeting her with harmful advertisements, connections, and subject matters. Shortly after the pandemic began is when N.M.'s parents noticed something was wrong. N.M. was home more due to quarantining and was engaged in more physical activities than usual. At first, her parents thought it was boredom, but then they noticed that she was losing weight and no longer enjoyed social situations. Specifically, she began focusing on calories, limiting the times she could eat, and asking for special foods with a precise degree of knowledge she should not have had. N.M. was being fed an arsenal of very specific information about foods, eating, and exercise, and despite her parents' diligent efforts to find out the source, they could not. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] To N.M.'s parents it felt like someone else had come into the picture and was significantly influencing their child who was changing before their eyes, but they could not figure out who, how, or why. This is one of the greatest harms Meta and TikTok have caused to users, families, and medical providers. Specifically, they concealed the harms only they knew their products were causing, which harms related to their product design decisions, programming, distribution, and operational choices within their sole knowledge and control. N.M.'s parents could have intervened and obtained N.M. the help she needed sooner and before the worst of the damage was done had these companies simply come forward and told users and their parents the truth. In September of 2021, N.M.'s mother, E.M., took her to the doctor to get checked out. The doctor said that something was wrong. N.M. was not getting enough calories, but that it was not out of control and her parents needed to start supervising her at meals. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] N.M. was then sent to an eating disorder specialist in the practice three days later, and N.M.'s parents noted the sharp decrease in her BMI, which coincided with the period of increase in her use of the TikTok product. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] E.M. then made the difficult decision to pull N.M. out of the private high school she was attending, so that she could better monitor and help her daughter get better, including by being able to bring her home at lunch each day to eat. This was the high school N.M. and her parents chose because of the opportunities it offered for her future, including a global exchange program through which N.M. planned to spend a semester in Ireland and specialized introduction to the medical career courses N.M. wanted to take in preparation for becoming a doctor. N.M. also participated in a field hockey league in preparation for a college scholarship and eventually she had to quit the league and competitive playing due to harms caused to her by Meta and TikTok. Unfortunately, N.M. did not get better. What E.M. did not know and had no way of knowing at that time was that Instagram and TikTok were causing the harms to her daughter. N.M. was dependent on their products and that they were fostering, connecting, and enabling harmful connections for their own gain. Instagram and TikTok were targeting N.M. on the basis of gender and other perceived vulnerabilities, while at the same time marketing themselves as fun and safe products for teens. Instagram and TikTok knew about the harms their products and programming decisions were causing and made the calculated business decision to keep the rest of the world in the dark. Millions of American children, including N.M., were harmed as a result. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In October 2021, N.M. began having chest pains. E.M. took her to urgent care, thinking it was a muscular issue. Instead, urgent care rushed N.M. to the hospital where she was admitted for a week with bradycardia and anorexia nervosa. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Shortly after N.M. was admitted to the hospital, E.M. went home to grab some things and quickly returned. When she returned, N.M. was in tears. She asked her mother how her phone knew what she was doing and said that she felt like someone was watching her. Then she showed her mother her TikTok For You Page and the videos TikTok had only just started aiming at her--videos that were focused on how she could convince doctors that she was eating and that she did not need to be sent away for treatment. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This included things like encouraging her to drink water before a weigh-in and other tips and tricks. N.M. had never received these videos on the TikTok product before, but TikTok began sending them moments after she was admitted to the hospital and left in her hospital room. TikTok and Meta collect thousands of data points on every user, from device ID to browser history, IP address, estimated age, education, consumer preferences, health history, and other categories that are incredibly invasive and inherently harmful--particularly given the manner in which these Meta and TikTok choose to use that information. While it is unknown how precisely TikTok determined that N.M. was in the hospital and seeking treatment for her eating disorder, TikTok knew and decided to target her with specialized subject matters aimed at helping her hide the eating disorder TikTok helped cause in the first place. All N.M. could think about was being released so that she could lose the weight she was forced to gain. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] N.M. began intensive therapy but was hospitalized again three months later, in April 2022. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This time, she was admitted to a facility that made clear things were now out of her parents' hands and that the level of care she would need if she did not make more progress would require a rehabilitation facility. N.M. committed to making more progress and was allowed to return home, where she worked to stay in recovery. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] As a result of her compulsion to interact with the Instagram and TikTok products, and specifically due to recommendations and content Instagram and TikTok selected and showed to N.M., a minor user of their products, N.M. subsequently developed injuries including, but not limited to, social media compulsion, sleep deprivation, anxiety, depression, body dysmorphia, anorexia nervosa, and significant hair loss among other harmful effects, which may cause or contribute to additional harms. Meta and TikTok designed their products to promote social comparison and body dysmorphia by providing beauty filters that encouraged users to present as having slimmer faces, smoother skin, bigger eyes, larger lips, and more. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta and TikTok designed their products to frustrate parents like E.M. from exercising their rights and duties as parents to monitor and limit their children's use of Meta and TikTok' products. Meta and TikTok designed their products to allow minors to use, become addicted to, and abuse their products without the consent of the users' parents, like E.M. Meta and TikTok designed their products to be attractive nuisances to underage users but failed to exercise the ordinary care owed to underage business invitees to prevent the rampant, foreseeable, and deleterious impact on minor users that access Meta and TikTok' products. N.M. and E.M. were not aware of the addictive and mentally harmful effects of Meta and TikTok' Instagram and TikTok products when N.M. began using those products. Meta and TikTok not only failed to warn them of the dangers of social media compulsion, sleep deprivation, and problematic use of Meta and TikTok' products, but misrepresented the safety, utility, and non-addictive properties of their products. Executives at both Meta and TikTok made public statements about the safety of their products and testified under oath before the U.S. Senate as to the same even though they knew or should have known of the harms their products were causing. As a result of N.M.'s extensive and problematic use of the Instagram and TikTok products she developed numerous mental and physical health conditions. The long-term physical harms caused by her anorexia remain unknown, but she has been informed by doctors that it is possible her long-term health will suffer, including but not limited to things like bone density, her ability to have children, and heart health. N.M. lost large amounts of hair that still has not grown back. She still suffers from anxiety, depression, anorexia nervosa, and social anxieties she did not have before her use of Meta and TikTok' products. N.M. attends college but her dreams of attending a top tier college with a sports scholarship were destroyed when Meta and TikTok weaponized their products and exploited her for their own gain. She no longer plays field hockey and still struggles with addiction to Instagram and TikTok. j. Patterson v. Meta et al. On May 14, 2022, 18-year-old Payton Gendron entered Tops Friendly Markets supermarket on the East Side of Buffalo and murdered ten Black citizens in a heinous act of racist brutality. As Gendron's attorney acknowledged, ``the racist hate that motivated this crime was spread through on-line platforms.'' Gendron did not grow up in a prejudiced household and did not hold racist beliefs until he began using and became dependent on social media. Gendron wrote that ``when I was like 12, that was when I didn't dislike American blacks and liked listening to black music.'' Gendron became addicted to social media products when he was a minor under his parents' legal control accessing his social media accounts multiple times per hour and at all hours of the night. He did not initially seek out racist material, but because social media apps were expressly designed to maximize engagement over his psychological and ethical well-being, they foisted upon him unsolicited material promoting racist hate and violence. Exploiting Gendron's incomplete frontal lobe development, they targeted him with increasingly extreme and violent content and connections which promoted racism, antisemitism, and gun violence. This was no mere accident or byproduct of poor software engineering; rather, these algorithms were functioning as designed and intended. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The material force-fed to Gendron progressed from white supremacist screeds to livestream videos of mass shootings and other extreme depictions of racist violence. This material inspired Gendron to move forward with his murderous plan. He explained: Knowing that so many other attackers like myself are out there rooting for me gives me quite a bit of confidence. Every single White man has everything to lose by doing nothing, and everything to gain by taking action. Yes I do find inspiration from other attackers. Gendron's ability to livestream his racist murder furnished the final motivation to carry out his evil plan. With haunting insight, Gendron explained how Christchurch shooter Brenton Tarrant's livestreaming of his March 2019 massacre increased the power of his racist message exponentially over the written statement released by Charleston shooter Dylann Roof: Dylann Roof's manifesto is not that bad Livestreaming this attack makes a 1000x greater impact I most likely wouldn't even know about the real problems in the world if Brenton Tarrant didn't livestream his attack. Gendron wrote that livestreaming the attack would help him overcome his fear and any lingering sympathy for human life that could dampen his murderous intent. It is very difficult for a normal person even with all the information to carry out an attack that will kill another human being, or the fact that you may die that day I think that live steaming this attack gives me some motivation in the way that I know that some people will be cheering for me. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The New York Court sentenced Gendron to life without parole, and he currently faces Federal death penalty charges. At his sentencing hearing, he acknowledged the web-based source of his radicalization: I cannot express how much I regret all the decisions I made leading up to my actions on May 14th. I did a terrible thing that day. I shot and killed people because they were black. Looking back now, I can't believe I actually did it. I believed what I read on-line and acted out of hate. Gendron's rampage was neither an accident nor a surprise. Teenagers' incomplete neurological development makes them ideal targets for Respondents' addictive algorithms and highly susceptible to radicalization. In the case of young male adolescents such as Gendron, social media platforms' addictive features (autoplay, infinite scroll, notifications, and social validation features including ``likes''), when paired with engagement-maximizing algorithms, actively encourage, assist, and facilitate the spread of racist, antisemitic, and terrorist propaganda. Taking full advantage of the incomplete development of Gendron's frontal lobe, Instagram, YouTube, and Snapchat maintained his product engagement by targeting him with increasingly extreme and violent content and connections which, upon information and belief, promoted racism, antisemitism, and violence. Because the Instagram, YouTube, and Snapchat algorithms were designed with the singular goal of maximizing Gendron's product engagement over his psychological, emotional, and ethical well-being, they directed him to sites and other users promoting hate and violence. As evidenced by this internal Meta document, these products were functioning as designed and intended. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The social media companies and streaming companies that Gendron used not only could have foreseen the significant radicalization risk posed by their apps--they actually saw the risk. Meta personnel aware that its recommendation algorithms ``are prone to recommending harmful content.'' In one experiment from 2019, found that in just 3 weeks by following just this recommended content, the test user's News Feed had become a near constant barrage of polarizing content, misinformation, and violence. Google recognized, anticipated, and even rebuked internal attempts to mitigate harms to their young users. Facebook's Director of Content Policy explained that not ``not only did we anticipate murders and suicides on Live, we anticipated far worse (all of one of our top 5 predictions have played out).'' When it acquired Twitch, Amazon knew that the product was used by criminals to livestream criminal activity and that the ability to livestream acts of violence using Twitch's product motivates criminals to follow through with their plans. R. First Amendment Concerns a. Companies Need Not Remove Content to Design Safe Products ``At the heart of the First Amendment lies the principle that each person should decide for himself or herself the ideas and beliefs deserving of expression, consideration, and adherence.'' These words, written by Justice Kennedy over 30 years ago, represents the bedrock essence of the First Amendment. Importantly, the protection applies both to speaker and listener: it defends not only the right of the speaker to be heard but also the right of everybody in the audience to listen. The First Protection is a most important repudiation of the censorious instinct of those in power, taking refuge in the false security of consensus, to stifle the voices of those with whom they disagree. What it does not protect, however, is misconduct. Social media companies regularly insist that they are immune from liability under every possible circumstance: that Section 230 protects them from liability arising from third-party content while everything else associated with their platforms should be treated as first-party speech. But as the ongoing, nationwide litigation against these companies has demonstrated, these companies can manifestly conform their conduct to the standards of a reasonable business without removing, editing, deleting, or moderating a single piece of content. History has consistently shown that censorship by the state is never the panacea so envisioned by those in power. The challenge before us is that emergent and miraculous technologies run into human tendencies as old as civilization itself: greed, power, ego, and control. Our ingenuity at designing social media platforms, complex algorithms, and artificial intelligence is despoiled by the inexorable need to maximize profits over safety or to serve other ignoble ends. We simply cannot help ourselves, and so the law must step in as a bulwark to protect society from the excesses and appetites of the few. Let me sharpen the point plainly: Congress should take no action to remove or moderate lawful content from any social media platform, no matter how deplorable, reprehensible, or offensive it may seem. After all, esteemed and learned Senators: to whom shall you award the right to decide which speech is harmful and who is the harmful speaker? To determine in advance for you what you can read, to relieve you of the responsibility of hearing what you might not wish to hear? What Congress should endeavor to do--and what my law firm has fought to do over these past four years--is hold social media companies (indeed, all companies) to the same standard of reasonable care and safe product design as every other industry. Such bedrock standards of care, referred to as a ``common law'' duty, are fully compatible and consistent with our first and most important Amendment. b. The First Amendment Protects Speech, not Conduct It is axiomatic that the First Amendment protects speech, not conduct; but crucially, it is also true that conduct is not automatically shielded by the First Amendment simply because it involves speech. Federal case law makes clear that generally applicable laws do not offend the First Amendment simply because their enforcement has an incidental effect on speech. When a bookstore sells a book, there can be no liability for the substance and content of that book. But when a publisher targets your children, constructs a bookstore like a casino, interrupts your children's school and sleep hours with notifications of new books, and deluges your children with a force-fed feast of psychologically discordant material for the express goal of ensuring your children never leave--in these circumstances, the First Amendment offers no refuge for outrageous misconduct. Throughout the country, in lawsuit after lawsuit, SMVLC has sought to hold social media companies accountable to a civil jury for the design of their products: the decision to include intentionally addictive platform features like infinite scroll and engagement- optimized algorithms, the manipulation of social validation principles such as ``likes'' and ``FOMO'' to hook children, and the use of ``dark pattern'' designs that frustrate the ability of both users and parents to realize the danger and take steps to protect themselves. Neither these lawsuits nor any act by Congress should seek to limit any platforms' capacity to engage in content moderation--to filter, prioritize, or label various messages, videos, or other content their users wish to post. Indeed, presenting third-party content is a key component of social media, but how they do so is the key issue. c. AI-Driven Algorithms and Chat Bots are Not Speech ``Speech is expressive. Speech communicates ideas. Speech has a message even when the message is not clear or is open to interpretation.'' This observation, made last year by District Court Judge Anne C. Conway in Garcia v. Character Techs., Inc., encapsulates why artificial intelligence is not entitled to First Amendment protection. Simply put, the First Amendment protects the freedom to think and speak as an inalienable human right. Indeed, leading scholars have explained the deeply concerning consequences of assuming machine speech is legally equivalent to, and deserving of, the same constitutional rights as human speech.\4\ --------------------------------------------------------------------------- \4\ See Tim Wu, Machine Speech, 161 U. Pa. L. Rev. 1495, 1496 (2013); Helen Norton, Manipulation and the First Amendment, 30 Wm. & Mary Bill Rts. J. 221, 223 (2021). --------------------------------------------------------------------------- By contrast, the ``editorial decisions'' for which social media companies seek constitutional protection consist of electronic impulses of artificial intelligence-driven algorithms designed to exploit the neurological immaturity of minor users and maximize their engagement by deluging them with psychologically discordant material that they are not seeking but from which they cannot look away. These artificial- intelligence driven algorithms are software, lines of code designed and written by software engineers to respond to specific variables without any underlying cognitive process or communicative intent. And once designed, these algorithms go forth and act in accordance with their marching orders and design parameters. In this regard, social media companies exercise no editorial discretion whatsoever--they designed recommendation algorithms that act based on data analytics and user metrics, not upon any editorial judgments by the companies as to desirable or undesirable content. Their algorithms could be designed to optimize a user's content recommendations for happiness, optimize for wellness, optimize for education--or, in the case of social media companies, optimize for screen time and engagement. Of course, there is also the point of what happens when social media companies let their AI algorithms off the leash. As Justice Amy Coney Barrett explained: Consider, for instance, how platforms use algorithms to prioritize and remove content on their feeds. Assume that human beings decide to remove posts promoting a particular political candidate or advocating some position on a public-health issue. If they create an algorithm to help them identify and delete that content, the First Amendment protects their exercise of editorial judgment--even if the algorithm does most of the deleting without a person in the loop. In that event, the algorithm would simply implement human beings' inherently expressive choice ``to exclude a message [they] did not like from'' their speech compilation. But what if a platform's algorithm just presents automatically to each user whatever the algorithm thinks the user will like-- e.g., content similar to posts with which the user previously engaged? The First Amendment implications of the Florida and Texas laws might be different for that kind of algorithm. And what about AI, which is rapidly evolving? What if a platform's owners hand the reins to an AI tool and ask it simply to remove ``hateful'' content? If the AI relies on large language models to determine what is ``hateful'' and should be removed, has a human being with First Amendment rights made an inherently expressive ``choice . . . not to propound a particular point of view''? In other words, technology may attenuate the connection between content-moderation actions (e.g., removing posts) and human beings' constitutionally protected right to ``decide for [themselves] the ideas and beliefs deserving of expression, consideration, and adherence.'' So the way platforms use this sort of technology might have constitutional significance. Moody v. NetChoice, LLC, 603 U.S. 707, 746 (2024). Certainly, algorithms could be designed to act upon specific editorial judgments, such as to rank or prioritize certain types of content over others based on the substance of that content (e.g., emphasizing pro-Ukrainian content over pro-Russian content or pro- Israeli content over antisemitic content). In fact, both President Trump and Congress previously recognized the power and danger of content-based algorithms in the hands of massive tech companies. In August 2020, President Trump issued an Executive Order finding that ``the spread in the United States of mobile applications developed and owned by companies in [China] continues to threaten the national security, foreign policy, and economy of the United States.'' President Trump determined that TikTok raised concerns, noting that the platform ``automatically captures vast swaths of information from its users'' and is susceptible to being used to further the interests of the Chinese Government.'' Against this backdrop, Congress passed the Protecting Americans from Foreign Adversary Controlled Applications Act (PAFACA), which sought to respond to concerns that TikTok--whose parent company is subject to the control of China--could use its sophisticated algorithm to harm U.S. national interests. Lawmakers recognized that a foreign adversary could use content-recommendation algorithms to emphasize and spread polarizing, misinforming, or otherwise destabilizing content throughout the U.S. The Supreme Court rejected TikTok's challenge of the Act under the First Amendment, holding that while ``laws favoring some speakers over others demand strict scrutiny when the legislature's speaker preference reflects a content preference,'' such scrutiny is ``unwarranted when the differential treatment is `justified by some special characteristic' of the particular [speaker] being regulated.'' TikTok Inc. v. Garland, 604 U.S. 56, 72 (2025). In sum, the social media companies' artificial intelligence algorithms cannot be fairly characterized as constitutionally protected ``editorial judgments;'' they serve no communicative purpose, are not ``speech,'' and are not entitled to First Amendment protection. d. Not All Speech is Protected Speech Even if social media algorithms and artificial intelligence chatbots are to be considered expressive speech, the question remains whether they depict protected speech. The First Amendment historically permits restrictions upon the content of speech in several limited areas. For example, it has never been deemed an abridgement of free speech to make a course of conduct illegal merely because the conduct was carried out by means of language. Speech integral to illegal conduct--e.g., defamation--cannot excuse the wrongful conduct underpinning it. Another example includes speech obscene to minors. On this point, little elaboration is necessary: the First Amendment does not protect adults from preying on and directing sexual communications toward minors. Yet time and time again, social media platform features and recommendation algorithms make it much easier to connect children directly to adult predators. And there have been countless instances of AI chatbots engaging in, and promoting, sexually explicit ``conversations'' with minors. This is no mere thought experiment or future problem. This ever-present danger to children is here among us right now. And finally, there has always existed some categories of speech that have been historically unprotected yet have not yet been specifically identified or discussed in the case law. A growing number of scholars have called for an exception to the First Amendment for ``manipulative expression,'' which functions more akin to content than speech. Indeed, the First Amendment already recognizes a special category for commercial speech, motivated by the impact of comparatively powerful speakers' comments on comparatively vulnerable listeners' rights. The recognition within the First Amendment of the disparity in power and information between speakers and listeners in certain circumstances, and the resulting frustration of listeners' rights, compels carving out efforts to regulate manipulative speech. Thank you all again, Mr. Chairman, for the opportunity to testify before this Committee and I look forward to answering your questions. The Chairman. Thank you, Mr. Bergman. And thank you for telling your client's story. And I want to say to the moms and dads who are here, thank you for being here. Thank you for standing up and remembering and fighting for your kids. I will say, as the father of two teenage girls, you are living every parent's nightmare. And I don't know a parent of adolescents or teenagers who is not terrified of the tragic forces that targeted your children. So, thank you. Mr. Carson. STATEMENT OF BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION Mr. Carson. Chairman Cruz, Senator Schatz, and members of the Committee, thank you for inviting me here today to testify as we mark the 30th anniversary of Section 230. I also want to give special thanks to Chairman Cruz for his determined and prescient work to pass the Take It Down law, a law which ARI supported, and which will make a difference in the effort to hold Big Tech platforms accountable. As we discuss harms to children online in the wake of Section 230, the Take It Down Act is a commendable effort to correct the historical course. Thirty years ago, Congress passed Section 230 to address a narrow problem, whether online platforms could moderate content without assuming publisher liability. Today, the law is widely criticized for enabling unchecked harms online. As we confront artificial intelligence, we should ask ourselves a simple question, will we repeat the mistakes of Section 230, or will we learn from them? There are two competing interpretations of how Section 230 went so wrong. One view is that Section 230 was ill-conceived from the beginning, and that ordinary Tort and First Amendment case-by- case development through the common law would have produced a more nuanced and flexible body of law. Rather than allowing the development of rules through our legal system, however, Section 230 froze answers in place to questions the law had not yet fully considered, much less answered. Another view is that the statute was defensible, but courts interpreted it far beyond what Congress intended, providing platforms with legal immunity, not only for user content, but extending legal protections to include significant algorithmic amplification and product design choices. Regardless of which is more accurate, both assessments recognize that the resulting legal regime has been unable to hold platforms accountable in proportion to the harms they have enabled. Families whose children were exploited online or harmed by algorithmically amplified content have too often discovered that the law offers them little recourse. Section 230 may be the 26 words that created the internet, but they also are the 26 words that have visited irrevocable harm to generations of children. I recount the history of Section 230 because Congress now faces a structurally equivalent moment regarding artificial intelligence. Some have suggested freezing State law developments while enacting no comprehensive Federal regulation of the technology. Section 230 is what I would call a meta law. A law that determines who governs an emerging industry, rather than how that governance would occur. Enacted when the Internet was still in its infancy, Section 230s consequences have proven extremely difficult to correct once the industry matured. In practice, the very first meta law on technology often becomes the last major law, because it determines who has the power to resist future corrections. I should be clear that Section 230 should not be interpreted to immunize the provider of a generative AI system for harms caused by system outputs. Section 230 protects platforms from liability for content, as the statute says, provided by another information content provider. That framework assumes active users and passive hosts. Generative AI systems do not fit that model. A user provides a prompt, but the company designs the model, selects the training data, fine-tunes the system, and deploys it with parameters of its choosing. The resulting output is not third-party content. It is the product of the AI system that the company created. Even though that answer does seem obvious, Congress should clarify the law before lower courts extend immunity further than the text permits. The most relevant lesson of Section 230 is that Congress should not establish broad, essentially unchangeable immunity for an industry that it does not yet understand, and its future development can scarcely be charted. The Federal preemption of State laws related to artificial intelligence would be a repeat of Section 230s tragic history, replete with all the errors of the former and without any of the latter's corrective measures. There are many forms of legal stagnation that Section 230 has caused. I would urge the Committee not to repeat these same mistakes as we embrace new technologies yet again. Thank you for your time. I look forward to your questions. [The prepared statement of Mr. Carson follows:] Prepared Statement of Brad R. Carson, President, Americans for Responsible Innovation Chairman Cruz, Ranking Member Cantwell, and Members of the Committee: Thirty years ago, Congress passed Section 230 to address a narrow but critical problem: whether platforms could moderate content without assuming publisher liability. Today, that law is widely- and rightly- criticized for enabling unchecked harms. As we confront artificial intelligence, we must ask, ``Will we repeat the mistakes of Section 230, or will we learn from them?'' Some believe that Section 230 was ill-conceived from the very beginning and that ordinary tort and First Amendment case-by-case development through the common law would have produced a more nuanced and flexible body of law than did the broad statutory immunity of Section 230. On this view, Section 230 did, it should be acknowledged, accomplish one important task. Subsection (c)(2) of Section 230, which grants immunity to platforms for good faith content moderation decisions, directly addressed a real concern created by cases such as Stratton Oakmont v. Prodigy. In Stratton Oakmont, a trial court held that a platform's attempt to curate content exposed it to liability as a publisher, thus penalizing the responsible content moderation. In correcting this erroneous court ruling, the protection provided by Section 230(c)(2) was very reasonable. However, subsection (c)(1) of Section 230, the broad immunity for third-party content, was not necessary. The common law and First Amendment were capable of producing the rules governing liability for third-party content, and the courts could have developed those rules over time in response to actual harm caused. Rather than allowing the development of these rules to occur, Congress in Section 230 froze the existing answers to questions that the law had not yet fully considered. Another view is that Section 230 was defensible as enacted, however the courts interpreted it far beyond what Congress intended. On this view, the primary error occurred in the Fourth Circuit in Zeran v. America Online in 1997, where the court eliminated the long-standing distinction between publishers and distributors. Publishers are those entities that exercise editorial control over content and therefore bear full liability for the publication of that content. Distributors, on the other hand, are those entities that merely transmit content knowing its character and bear a limited liability based upon their role. Before the enactment of Section 230, a distributor that knew of particular harmful content and failed to take action against that content could be held liable for failing to act to prevent the distribution of the content. The Fourth Circuit interpreted Section 230 to eliminate that distinction altogether, holding that the statute granted platforms immunity from liability even when the platform had actual notice of the existence of specific harms and chose to take no action to prevent the continued distribution of the harmful content. Following the decision in Zeran, subsequent decisions extended that immunity further still, to include algorithmic amplification, product design, and to business decisions that had nothing to do with the hosting of user-generated content. The statute was not the problem, according to this analysis, but rather the interpretation of the statute went awry. Regardless of which is more accurate, both assessments of Section 230 recognize that the resulting legal regime has been unable to hold platforms accountable in proportion to the harms they have enabled. Families whose children are victimized by predators have been denied access to the courtroom. Parents whose teenage child experienced years of self-harm associated with foreseeable algorithmic targeting and engagement-optimization design choices discovered that the law too often offers them no realistic path to accountability. The social media platforms have borne few of the costs of the harms they have facilitated. Rather, those costs have fallen to the individuals that have sadly learned that the law affords them little protection when technology goes wrong. I recount the controversial history of Section 230 because Congress now finds itself facing a structurally equivalent moment regarding artificial intelligence. Many have suggested that we freeze state law developments while enacting no comprehensive Federal regulation of the technology. No matter whether the fault rests with the statute itself or with the interpretations that followed, the ultimate result of Section 230 was a meta-law--a law that determines who will govern an emerging industry instead of how that governance will occur--enacted for an industry that was in its infancy and whose consequences proved nearly uncorrectable once the industry matured. That is the error that Congress is being encouraged to repeat with respect to artificial intelligence. In practice, the first meta-law often becomes the last major law because it determines who has power to resist future corrections. I. Section 230 Should Not Be Interpreted To Immunize Generative AI Outputs I would like to make clear that Section 230 should not be interpreted to immunize the provider of a generative AI system for harms caused by the system's own outputs. Section 230(c)(1) grants immunity to platforms for the content ``provided by another information content provider.'' Platforms receive this immunity because the content posted by users originates with the user; the platform is merely the host, not the author. The statute contemplates a world in which there exist active users and passive platforms. The latter are generally immunized from the actions of the former. (Of course, when a platform merely hosts content posted by users--including content those users generated with AI tools--Section 230 may still protect that hosting function. But the company that designs, trains, and deploys a model should not be treated as a passive host of the model's own speech-like outputs.) A large language model does not provide third-party content in the way contemplated by Section 230(c)(1). A user provides a prompt to the model. The model, having been trained on a dataset selected by the company, having been fine-tuned by the company, and having been deployed with parameters determined by the company, produces the output. Therefore, the company is responsible, in significant measure, for the creation of the content. Pursuant to the statute's literal text, the immunity of Section 230 does not apply to artificial intelligence; the output of Claude or ChatGPT is not the ``information content'' of ``another.'' I believe, though, that Congress should clarify this before lower courts, following the Zeran precedent, extend immunity further than the text of Section 230 permits and establish an erroneous precedent that will require yet another generation to correct. Legislation introduced by Senators Hawley and Durbin would make clear that Section 230 does not apply to generative AI, and I would urge Congress to move swiftly to enact it. I ask Congress to act even though the textual answer seems plain. The lesson of Section 230 is precisely that textually plain answers do not always survive contact with courts under pressure to resolve novel cases quickly and in the absence of congressional guidance. Zeran was not an unreasonable decision given the legal vacuum in which the Fourth Circuit found itself. It was, however, a consequential one. Congress should not again create the conditions in which courts are forced to improvise and then spend a generation arguing about whether the improvisation was correct. II. Congress Should Not Create An Industry-Wide Immunity For An Emerging Industry That It Cannot Yet Predict While the question of whether Section 230 applies to AI can be clearly answered in the negative, the more relevant lesson of Section 230 is that Congress should not establish broad, essentially unchangeable immunity for an industry that it does not yet understand and whose future development can scarcely be charted. When Congress passed Section 230 in 1996, it could not have imagined that a few websites would soon control virtually all public communication, influence election results, and have market capitalization larger than many countries' entire GDPs. Indeed, in 1996, just 7 percent of Americans had access to the internet, Netscape was the dominant browser, Google did not exist, and Facebook's launch was eight years away. Yet Section 230 shielded an entire infant industry from legal responsibility. The immunity has been difficult to repeal as the infant industry has grown beyond anyone's imagination. In retrospect, creating broad immunity for an emerging industry should be seen as a category error disguised as prudence. We do not permanently exempt a child from the law because the child appears harmless today; we calibrate accountability to capability and impose it consistently as capability grows. That is what a responsible legal system does. Section 230 did not do that. III. Preemption Without A Federal Framework Would Repeat Section 230s Mistakes The discussion regarding Federal preemption of state laws related to artificial intelligence is worthy of separate treatment here, because it is not merely analogous to the historical development of Section 230; it is a repeat of Section 230s history, replete with the errors of the former, and without any of the latter's corrective measures. Section 230 created two types of legal stagnation. The first was interpretive. Because Section 230 created a broad immunity, courts never developed common law governing platform liability for actual harms. The second type of stagnation was legislative. Once Section 230 became entrenched, amending or replacing it required a political consensus that proved nearly impossible to assemble. State legislatures found their efforts preempted before they could be tested. For its part, Congress could not create new legal structures without first eliminating an existing statute that had become a ``third rail'' politically. As such, Section 230 remains frozen in place as technology continues to revolutionize the world. Similarly, a Federal immunity toward AI--which is what the preemption proposals offered in Congress to date really are--will lead to the same stagnation, but more rapidly, and for a technology whose potential implications are much larger than those of social media. Consider what preemption without a Federal framework actually accomplishes. It removes a state's authority to apply its tort law, consumer protection statutes, and civil liability regime to AI systems. It prevents plaintiffs from having access to courtrooms prior to the development of a Federal alternative to receive them. It informs courts that they are under no obligation to develop a body of common law regarding AI liability, while providing no alternative for them to apply. In short, it is a legal vacuum created by Congress to which neither the courts nor the legislative branch can respond. This is the Section 230 model, deliberately reproduced. If Congress wishes to preempt state laws governing AI, then the cost of that preemption must include a genuine Federal framework that replaces the space that the preemption vacates. This Federal framework could include, first, a flexible Federal structure that assigns liability standards to agencies based on categories of risk, allowing those standards to evolve as the technology does rather than locking them into statute. Second, the framework could help develop markets for independent AI auditing and safety certification, established through procurement preferences and safe harbors conditional on third-party review, to create accountability incentives in the marketplace. Third, the framework could insist on mandatory liability insurance calibrated to a system's capability and risk profile, compensating victims while creating pricing disciplines that accumulate with experience. None of these approaches require Congress to forecast the nature of AI in ten years. To the contrary, that is one of the advantages they possess over broad immunity. Such a Federal framework establishes accountability proportional to capability, rather than immunity inversely proportional to foresight. IV. Conclusion Section 230 was an understandable, if lamentable, case of poor foresight exacerbated by a failure to permit the law to evolve as the technology evolved. The people who have suffered the most as a result of that failure have been the families, the teenagers, and the individual victims who discovered that the law offers them nothing when technology harms them. We should not repeat that story with artificial intelligence. Meta- laws and broad immunities for emerging industries rarely age well, and the lesson of Section 230 is that they age worst of all for the people least able to protect themselves. Congress has better tools available, including frameworks that scale with capability, standards that evolve with the technology, and accountability that grows as power grows. The choice is between mechanisms that establish accountability proportional to capability and immunities that are inversely proportional to foresight. Thank you. I am happy to respond to your questions. The Chairman. Thank you to each of the witnesses. Let's start off for Ms. Keller and Ms. Farid Johnson. With the benefit of 30 years of hindsight, what is the worst or most offensive use of Section 230 as a liability shield that you have seen? Ms. Keller. So, I will give you two answers. One is about racial discrimination. The second is about Government censorship. So the first is a case called Vargas. That was the one decision out of a cluster of cases. And the general allegation was that Facebook was offering racially based targeting, gender based targeting, et cetera, for ads for housing and employment and credit, which is prohibited by Federal law. And what makes that a--not a 230 case, in my mind, is that the user uploaded something totally legal, just an apartment listing, and Facebook added the thing that made it violate the law, which was the racial targeting. So, that is my Exhibit A. And I should repeat, Facebook alleged that, you know, we don't know exactly what happened. But the second is a case called Sikhs for Justice v. Facebook. This was an American court plaintiffs alleging that Facebook had silenced them based on pressure from the Indian government. And the court ruled that under Section 230, the platform had the right to do that, which I think is doctrinally correct. But it is very sad, right? Section 230 exists to protect users. It is there to give a spine to the spineless if they want one. And having platforms back down when they could have done better is unfortunate. But thanks to Section 230, hopefully another platform can come along and do better. The Chairman. Ms. Farid Johnson. Ms. Johnson. Thank you, Senator. I am actually going to use the example of a case that I think demonstrates both some of the cynicism in seeking 230 protection when it is unwarranted, and that demonstrates as well the limitations of the provision. There is a case called Lemon v. Snap in which a bereaved parent sued the company that is the parent company of Snapchat for its allegedly negligent design of a speed filter. That speed filter encouraged users of the app to drive at excessive speeds. Snap responded to the complaint by seeking 230 immunity. And the appellate went to the appellate court. The appellate court noted that a platform can still face liability for its provision of content neutral tools, where the liability stemmed from the platform's own acts, such as designing, here in this case, and making available to users the speed filter and the corresponding reward system. And it wasn't about the content posted by the users. It was about the actions of the platform itself. The Chairman. So, let me ask a follow up. A key motivation for Congress passing Section 230 was to incentivize a flourishing marketplace of ideas online. As I explained in my opening statement, I want to find ways to curb Big Tech censorship so that we have more speech, not less speech in online platforms. For Ms. Keller and Ms. Farid Johnson, how could Congress constitutionally modify Section 230 to protect or incentivize more speech, not less? Ms. Keller. Well, here again, I have two answers. And the first is about jawboning and pressure from Government. The second is about empowering users and diversifying our options. So I will get through the first one, and you tell me if you still have time for a second one. So, I didn't love the Biden Administration pressure that was illustrated in the Murthy case. That said, having been on the receiving end of pressure like that, the tone in that record looked like the tone that Government officials from all parties sometimes take in high- handed demands that platforms conform to their content preferences. So in that case, the plaintiffs got unprecedented discovery and still couldn't establish any actual causation that led to their speech being suppressed. And this led to the Supreme Court issuing a really problematic ruling on standing that for the real victims of real jawboning in the future, it is going to be harder for them to get into court because of that. We are in an era right now of jawboning that is unprecedented in my lifetime. Brendan Carr is at it again. We see him pressuring news outlets now. We are seeing highly politicized enforcement at the FTC of laws that are supposed to be neutral consumer protection and antitrust laws. All of this leads to unprecedented vulnerability of our speech and communication to this kind of pressure, because as you pointed out, Chairman, as both of you pointed out, all of our speech is very dependent on these big private companies right now. So, you know, we are seeing this playing out on the screen in front of us with platforms. Maybe it is also taking place in back rooms. But you know, right on TV, we can see back in August 2024, President Trump threatened to jail Mark Zuckerberg. In January 2025, Zuckerberg very publicly announced changes to Facebook's and Instagram's speech policies to conform with the current Administration's preferences. He did this in this like hostage video. It was very strange. And also in January settled a $25 million lawsuit that Facebook was going to win and paid that out in a suit from President Trump. And when President Trump was asked if these changes were a result of his threats to jail a CEO of a platform, he said, probably. This is crazy. This should not be possible in today's America. The First Amendment should preclude it. The any work that you can do, or you perhaps are doing, to make it more possible to get into court and to get past the standing restrictions from Murthy, I think would be extremely important. But in the meantime, as I mentioned, Section 230 can give smaller platforms more of a spine. The Chairman. And Ms. Farid Johnson, you can answer briefly because our time has expired. Ms. Johnson. I would just say briefly, Senator, that the way we see the best way to amend 230 is to condition its protections on the transparency, limited data privacy, and interoperability requirements we suggest. The Chairman. Thank you. Ranking Member Schatz. Senator Schatz. Thank you, Chairman. Thank you to all of the testifiers. I just wanted to say, at the risk of getting the Chairman in trouble in his home state, he is doing an extraordinary job of leading this on a bipartisan basis, and also specifically on the question of jawboning. I think without his leadership, there would have been fewer guardrails on the current jawboning. But I also take your point, Ms. Keller, that the jawboning occurred in the previous Administration as well. And the fog of war, the urgency of protecting people from the COVID-19 pandemic does not entirely justify the way that Federal Government officials interacted with private sector companies. So I am just glad that we now have, in three dimensions, in real time, examples of both parties doing this. And so, it is no longer theoretical that the door swings both ways in Washington, and this is going to bite us all in the butt, and we have to fix it. We now know it. And so, I am just pleased that Senator Cruz has set the platform for us to work on this on a nonpartisan basis, not a bipartisan basis. Mr. Bergman, your lawsuits are interesting to me. I am not a lawyer, but I am particularly interested in the kind of expansive view of what the platforms view as 230 providing them a shield. And what I think is your theory of the case, which is, no, the AI, the design choices, all of those are a product and they are subject to regular product liability. And I would like you to kind of articulate how that distinction is made and whether there is anything that we can do statutorily to kind of constrain the regular 230 immunity. Someone uploads some content. They can't be sued for libel. That is fine. But now we are in an age where half the stuff on these platforms is AI anyway. A lot of it is harming kids. And it seems to me that we need to clarify either through the court system or through the legislature or both what we actually originally meant by 230. So, I am interested in your thoughts. Mr. Bergman. Well, yes. And I think Section 230 remains an important protection for online communication and for the free exchange of ideas. I think what is important is that Section 230 adhere to its original intent, which was to immunize publishing activity. And, you know, the impetus for this came from my mentor, Judge O'Scannlain, on the Ninth Circuit in Barnes v. Yahoo, where he established a distinction between where an individual is seeking to hold the company liable for traditional publishing activity, bad content moderation, having bad stuff online, putting bad stuff online, and that is and should remain subject to Section 230. But as the court held in Barnes, even if the same harm results from a different theory of liability--in the case of Barnes it was promissory estoppel. In the case of Lemon v. Snap, it was negligent design. That is a separate source, it is a separate duty, and that that should be allowed to proceed. Our cases are based on the premise that these platforms are products, that they are designed based on deliberate design decisions. They target children not with material that they want to see, but material that can't look away from. They take advantage of the fact, and they exploit the underdeveloped frontal cortices of young individuals, the need for FOMO, fear of missing out, the social anxiety that adolescents have, and they use intermittent reinforcement techniques and very highly sophisticated AI to addict them to their platforms. And the research is showing that it is actually physically addictive. So I think that it is possible--and we have seen this. We just completed a trial and are awaiting a jury verdict as we speak, where we brought a case to the conclusion of trial, where we were able to cleave that important distinction between content moderation, which is protected by Section 230, even if the platforms aren't doing a good job of that, and the deliberate design of defective products. Senator Schatz. Perfect. Now, I am always a little cautious about writing a statute, a new statute, if I think that the existing statute already provides the pathway. Because I don't want to stipulate to us needing to change the law if your essential point of view is that actually the law just needs to be interpreted properly. And I think you have got it right. But I would like your advice over time about whether there is anything we can do in terms of clarifying legislative intent, or to include modifying the existing statute? But again, I don't want to stipulate to the idea that the original law provides such a broad immunity to the platform. So, what are your thoughts about the need for the legislature to take action, or should we wait for the court cases to win their way? Mr. Bergman. Well, if we wait for court cases to work their way, more kids are going to die, so I think things have to happen. I think that clarifying the original legislative intent to encourage the development of technologies, which maximize user control over what information is received by individuals, to remove disincentives to the development and utilization of blocking and filtering technologies that empower parents to restrict their children's access, and to ensure vigorous enforcement of Federal criminal laws to deter and punish trafficking and obscenity, stalking, and harassment. I think that those--if this committee were to reaffirm those objectives, along with the very important objectives of preserving the free marketplace of ideas, I think we could go a long way. Senator Schatz. So, one final thought. I agree with almost everything you said. But Senator Cruz and I have a bill on a bipartisan basis, the Kids Off Social Media Act. And the basic idea behind that, besides the obvious, that there is no real use case for a 9-year-old to be on Instagram, or TikTok, or Snap, or whatever it may be, is also that this idea of empowering the user to kind of turn the dials and the settings is a fantasy. If you are a parent, that is a fantasy. You are not going to get access to your kids' phone all the time. The tech companies, the platforms will work around whatever statutory framework we have. And that is why we need a sort of there is no safe cigarette point of view as it relates to kids, as you are pointing out. Brains and bodies are not fully developed. So, kids off social media app. I don't have any objection to empowering users, but I do not think that is going to actually protect kids. The Chairman. And I very much concur with what Ranking Member Schatz has said, that the Kids Off Social Media Act is one. This committee has already passed it in a bipartisan way, and I hope we pass it on the floor of the Senate and get it to the President's desk for signature. I will tell as a brief story just how helpless parents are in dealing with this. Several years ago, our eldest daughter was in trouble, and we grounded her and took away her phone for a month, which as parents of teenagers will tell you, is truly a draconian punishment. Two weeks into it, Heidi gets a random e-mail from Verizon that didn't make any sense. So we went and dug down more deeply, and it turned out that before she had handed her phone over to us, she had taken the SIM card out of it and had gotten a burner phone and put the SIM card on the phone. And we went to confront her, and at age 14, she sat with her arms crossed in her room. She said, ``Dad, you said I couldn't have my phone. You didn't say I couldn't have my SIM card''. And I have to admit, I was both annoyed and really proud at the same time. But it does show just how completely outmatched parents are trying to keep up with teenagers with these issues. Senator Fischer. STATEMENT OF HON. DEB FISCHER, U.S. SENATOR FROM NEBRASKA Senator Fischer. Thank you, Mr. Chairman. The challenge before Congress is ensuring that we have both accountability and the environment for free speech expression online. And Section 230 was designed to meet both of those goals. Ms. Farid Johnson, could mandated transparency about content moderation decisions improve accountability without chilling speech? And what is--really, what is the line here between protecting free expression and systematic harassment online? Ms. Johnson. Thank you, Senator. As my institute has argued in various briefs in this topic, transparency requirements, including disclosure and reporting requirements, are generally upheld under the Zauderer standard, so long as they are not unduly burdensome with respect to speech. So what that means in kind of plain terms is that there can be laws that require the disclosure of factual and uncontroversial information from a platform about the terms on which a service is offered, for example, as what you are mentioning. And these can be considered Constitutional if they are not unjustified. So essentially what it comes down to is if Congress can craft a provision that demonstrates there is not an undue burden on speech, but also provides a transparency that you are hoping for, that could meet--that could pass scrutiny. Senator Fischer. Is the problem here that we are looking at--in the statute itself, or is it with how the courts interpret that? Ms. Keller, we will start with you. Ms. Keller. Sorry about that. I am trying to think through the how the courts interpret it piece, because right now the courts are all over the place. You know, there are rulings---- Senator Fischer. Do we need to drill down and be more specific in a statute, which is kind of contrary to what Congress usually does? Ms. Keller. I actually think that the courts are in the process of working toward some answers. So for example, on this question of design liability and figuring out like what you would consider to be the platform's own doing and distinct from user speech, there are a bunch of child safety cases going on, including in multi-district litigation, where courts are saying, well, maybe autoplay of videos counts as design, and that is not immunized, and it is--you know, it is something platforms could be liable for. Maybe sending notifications counts as design. Maybe infinite scroll counts as design. And so, they are moving toward answering these questions that I think actually would play out kind of similarly under the First Amendment and 230 also. Senator Fischer. Do you think we need to look at Section 230, and should it be able to protect platforms when their algorithms are really actively amplifying some really harmful content, or only when they host it? So, are they--do we have a distinction out there between hosting and creating? Ms. Keller. So, I think the first question is, are we talking about content that is in the lawful but awful category, as so much of the pro-anorexia, just like terrible content is. If Congress doesn't have the power to prohibit that, they also don't have to power to tell platforms to change their algorithms and down rank it. The Supreme Court has been very clear in saying that a direct ban on speech and something that just burdens speech by making it harder to distribute get the exact same scrutiny from the court. Senator Fischer. If we would carve out algorithms out of Section 230, does that create a workable law, or is it just going to create a lot of litigation? Ms. Keller. So, this is the question that went to the Supreme Court in Gonzales, and then didn't get resolved. So there are a stack of briefs of very smart people arguing every side of this issue. But what I think is that it would cause the most important real estate on platforms, the places that most users actually go, to be Disney-fied, you know, to be purged of anything with any risk. For example, says, at the time of that case, at least 70 percent of YouTube views were from the recommended videos. And it would mean that in that area, YouTube has an incentive to definitely not have any, you know, anti-ICE videos, not have any accusations of wrongdoing by powerful people, not having new voices breaking in and have that be extremely safe. And I think that would be harmful for the many, many people, creators, advocates, who rely on platforms to get their word out. Senator Fischer. OK, thank you very much. Thank you, Mr. Chairman. The Chairman. Thank you. Senator Klobuchar. STATEMENT OF HON. AMY KLOBUCHAR, U.S. SENATOR FROM MINNESOTA Senator Klobuchar. Thank you very much, Senator Cruz, and thank you for holding this important hearing. And of course, our work together on the Take It Down Act couldn't be more important. I think I will start with you, Mr. Bergman. Can you talk about how the just blanket Section 230 immunity has shut the courthouse door to so many parents seeking justice for their children? Mr. Bergman. Yes. We have a case involving a 12-year-old girl who through Snapchat's friendship recommendation algorithm was connected to an online sex predator who was able to utilize the Bitmoji process to groom this child into sextorting her, meeting her, and he raped her. And we brought suit against Snapchat saying that the technology that allowed an unprotected--that failed to provide protection to a 12-year-old girl and allowed a sex predator to operate at scale and disguise his malevolent intent was a design decision. That case, unfortunately, was struck down or was dismissed. The court said, it is offensive to our conscience, but I have to do it. Senator Klobuchar. And how would repealing it or making changes to Section 230 create real incentives for platforms to be designed in a way that are safe for users? I just think of like an exploding washing machine. That wouldn't be protected from immunity. So, you know, they get sued, and then they make changes. Mr. Bergman. Senator, that is the exact point. We just want social media companies to follow the same incentive structure that every other company does. It is reasonable care. Senator Klobuchar. Mr. Carson, should we continue to think of these companies as merely neutral distributors, as they would like, as opposed to publishers picking and choosing what users view and when? Because the whole idea originally was distribution. Mr. Carson. I think the idea that they are in neutral platforms is normatively persuasive to me but runs Constitutionally afoul of how the Supreme Court approaches First Amendment law these days. I do think in kind of an answer to your question and to what Senator Fischer said, a great reform of Section 230 would be to remove the algorithmic optimization that the platforms do from being seen as another's speech. So right now, it doesn't protect your own speech, only another's speech. So if you were to define, say, what Facebook does to their optimization programs as Facebook's speech, it would still have First Amendment protections, and you would have to litigate those kind of questions. They would no longer be--it would be their own speech, and therefore not subject to 230 immunity. So there is a way to get after what is a very serious problem, which is the use of algorithms, which the Supreme Court has suggested does have some Constitutional protection. But these ones are just about optimization, that don't actually have any expressive value. I am not promoting speech about different causes, right. Senator Klobuchar. I understand. So, could the Take It Down Act--just as we look at beyond Section 230, is there some model we could use? We have the No Fakes bill to protect people's identities on the AI front and the like. Is that some kind of model we could use? Mr. Carson. That is the right model, the Take It Down Act. The DMCA, Section 512, similar, are both, I think, promising avenues. Senator Klobuchar. OK, very good. Thank you. Professor Farid Johnson, we need these commonsense rules of the road. And Senator Grassley and I have the American Innovation Choice Online Act, which we actually got through the Committee with a very good vote, to pry open competition online as another way to look at this by stopping anti-competitive self- preferencing. In your testimony, you argue that Congress should attack the platform's monopoly control over public discourse. Can you talk about how increasing competition and choice online, especially through interoperability, can help address some of the problems that we are talking about today? Ms. Johnson. Yes, Senator, thank you. What we are hoping to do here with this type of legislation is to give users more control over their experience. Right now, what happens is when someone logs on to, let's say, Facebook, for example, they are given a feed, and that feed is what comes to them because of the information that Facebook collects about them and Facebook is making recommendations accordingly. If somebody wanted to log on to Facebook and be able to talk with certain friends, or their grandmother, or whatever it might be, but doesn't want to deal with all of that, if we had a robust interoperability mandate, that person would be able to use third-party software called middleware to be able decide how they want to view that feed. So, it is one example of it. Senator Klobuchar. OK. Could I just go to you quickly, Ms. Keller. You previously said that middleware, that Ms. Farid Johnson just referred to, software can give users more control over what they see online. Talk about one solution created that could help with Section 230, but almost no major platform allows it. How would opening up these platforms mitigate online harms? Ms. Keller. I am so glad you asked because I actually, I wish that Senator Schatz were still here to respond to his very valid concern that parents don't want a bunch of choices, they just want their kids to be safe. Part of what middleware would make possible, or interoperability, is competitors who have figured out a better way to protect children to offer a one- click option for parents. Bringing in interoperability is a way for innovators to find new and different ways to do these things. So, I think it is extremely promising. Right now, people are scared to build interoperable projects because they can get sued and shut down. This happened to a company called Power Ventures 10 or 15 years ago. They offered a very basic service to aggregate your social media feeds from multiple platforms in one user interface, and Facebook sued them under the Computer Fraud and Abuse Act, which is this anti-hacking law from the 80s, and got them shut down. Similarly, Section 1201 of the DMCA, which is intended to prevent hacking through encryption on copyrighted DVDs, gets weaponized this way. So, I think there are a lot of opportunities to reform existing laws in ways that open the door to interoperability. Senator Klobuchar. OK, these were great answers. Maybe I will do a few more on the record, but thank you all for your thoughtful work, and I really hope we can press ahead with some of these changes. Thank you, Senator Cruz. The Chairman. Thank you. Senator Schmitt. STATEMENT OF HON. ERIC SCHMITT, U.S. SENATOR FROM MISSOURI Senator Schmitt. Thank you, Mr. Chairman. Ms. Keller, I assume as a First Amendment advocate you would agree that the Government shouldn't be censoring American speech, correct? Ms. Keller. Correct. Senator Schmitt. And the Government can't outsource that censorship to private actors either, can it? Ms. Keller. I hope not. Senator Schmitt. Right. Now, you are the Director of the Stanford Cyber Policy Center, which oversees the Stanford Internet Observatory, correct? Ms. Keller. No, I am not the Director. Senator Schmitt. OK. Do you have a role in that organization? Ms. Keller. I have a role in the organization, but my main placement now is in the law science technology program at the law school. Senator Schmitt. OK. Did you have anything to do with that observatory participating in the Election Integrity Project of 2020? Ms. Keller. Not really, but I fully support their First Amendment right to do so, you know. Senator Schmitt. Well, let's walk through that. So their role with the Biden Administration was to flag content that they thought was offensive, that may have--didn't line up with the Biden Administration's view of the origin of the COVID virus, or on the vaccine, or on people's opinions on how the 2020 election went down. And so, their--what their job was to do, which was to flag content that the Biden Administration didn't like, feed it back to the Biden Administration, and then put pressure on social media companies to take that content down. Is that consistent with your view of the First Amendment? Ms. Keller. No, I haven't--excuse me, that version of the facts is not consistent with my understanding. I don't think they were doing the Biden Administration's bidding. Senator Schmitt. What were they doing? Ms. Keller. They were exercising their First Amendment rights to go talk to the Government and say what they thought should happen. Senator Schmitt. What is it that they thought should happen? That content should come down? Ms. Keller. Probably, in some cases, as is their First Amendment right. Senator Schmitt. Are you defending that, that the Government should take that content down and put pressure on social media companies? Ms. Keller. There is no Government in this scenario. Senator Schmitt. Oh, there is. Absolutely. The Government outsourced to your university the job of flagging offensive content that they didn't like. And in this country, the Government doesn't get to decide what people see, or they hear, or what they get to say. And your university was right in the middle of all of this. And what I hear you saying is that you actually think that is OK. Ms. Keller. I think--again, academics exercising their First Amendment rights to say what they think and say it to Government are absolutely protected by the First Amendment. Senator Schmitt. That is not what was happening. That is not what was happening. The Biden Administration charged them with the duty of, we can't do this all ourselves. We wanted to start a disinformation governance board, but it got so much pushback, we can't do that. So we are outsourcing to Stanford to flag the stuff that the Government agency won't do now because we don't have one. Tell us what is offensive by way of these guidelines, and then we will tell the tech companies, that is a nice tech company that you have got there. It would be a shame if something happened to if you don't throttle these people, or take down RFK Jr., or take down Jay Bhattacharya, who is now the head of NIH. So I find it a little rich that you are here lecturing us about and espousing the virtues of the First Amendment, and then on the other hand defending your university's role in the censorship regime that we lived through for four years. Ms. Keller. I wasn't there. I can't tell you how those conversations went. But I do know the people---- Senator Schmitt. I wasn't there, and I can't---- Ms. Keller. I can't imagine them doing the Government's bidding. Senator Schmitt. Well, I can tell you can read all about it in Missouri v. Biden, the lawsuit that went to the Supreme Court. Ms. Keller. The one that you lost? Senator Schmitt. No, it is actually back. It was sent down to lower court for standing issues. So, we didn't lose the case. And I think it is telling--I actually have no idea why you are here today. You have embarrassed yourself. Your university has embarrassed itself. It is part of this censorship regime, which by the way is a cautionary tale for the future. And so, yes, it is a cautionary tale. So Mr. Bergman, I want to ask you. Since Section 230 was enacted, are there features that exist now that weren't contemplated maybe with Section 230 that are worth taking a look at? I happen to believe having an open platform for people to share their points of view is very important in this country, but there have been obviously outcomes that are terrible that we are talking about today. Are there certain things that protecting that open platform for people to add that pressure release valve to speak their mind, but especially as it relates to kids, or other things that maybe weren't contemplated in the early or the late 1990s that we could address? Mr. Bergman. Absolutely. You know, Netscape was the biggest online platform when Section 230 was enacted. But yes, Senator, there are specific features, the infinite scroll, the like feature, the streaks, the push notifications that are designed to addict kids. And again, not by showing them what they want to see, but what they don't want to look away from. If a 12-year-old girl really wants to access anorexic content, God forbid, that is a very sad situation, but I don't think it gives rise to liability. On the other hand, if the platforms, only in order to maintain an addictive relationship to sell more ads, feed that child information she is not looking for, I think that is a distinction that can be drawn and can preserve the vibrancy of the Internet as a free marketplace of ideas. Senator Schmitt. Thank you. Mr. Chairman, I would just point out that I filed the COLLUDE Act, which would say basically that if you are a social media company, and you get into the content moderation business and you violate somebody's First Amendment rights, there would be a private right of action, and you would lose your Section 230 protections. I think that would be a very important reform for us to consider. Thank you. The Chairman. Thank you. Senator Baldwin. STATEMENT OF HON. TAMMY BALDWIN, U.S. SENATOR FROM WISCONSIN Senator Baldwin. Thank you, Mr. Chairman. So jawboning is informally defined as informal, often coercive efforts by Government officials to pressure private companies into moderating or removing content that they cannot legally censor directly. I was listening with interest of the previous Senator's questioning, and certainly feel like there is examples we can draw upon from multiple Administrations, including this one. President Trump has attempted to rewrite history by forcing museums to remove content. Brendan Carr, the head of the FCC, has threatened broadcasters' licenses, including very recently who air unflattering news about this Administration and this President. If used properly and validated by experts, the Internet can be a place that people turn when information is being limited or attempted to be censored by the Government. Think about now and recent events. It might be where somebody accesses information on abortion, or LGBTQ identity, or scientific research on climate change, or your rights against discrimination in the workplace. So, I want to talk about the upsides. I also want to talk about the downsides of Section 230. But Ms. Keller, can you explain whether and how Section 230 works to protect access to information that people rely on to make informed decisions? Ms. Keller. Section 230 is essential in protecting that access. You referenced information about reproductive health care and abortion. Under current Texas law and under a bill that I believe is still pending there, it would be very easy for someone to sue a platform because a young woman researched this information. So, I think you are completely right to connect those two issues. Senator Baldwin. Mr. Bergman, your testimony and the work that you do clearly highlight the harms that can come from the Internet and technology, especially for children. I want to thank the parents who have brought their stories and their children's stories here with them today and are actively taking on these social media companies. Your stories clearly highlight the immediate need to remove harmful content, especially content directed toward children from these sites. So I would like to just, first of all, go down the line and ask all of the witnesses, does Section 230 prohibit platforms from engaging in content moderation? Ms. Keller? Ms. Keller. No. Senator Baldwin. Ms. Farid Johnson? Ms. Johnson. No. Senator Baldwin. Mr. Bergman? Mr. Bergman. It does not. Senator Baldwin. Mr. Carson? Mr. Carson. No. Senator Baldwin. All right, so Ms. Farid Johnson, in your testimony, you highlight that Section 230 does not protect platforms from harmful design decisions they make. When is a design decision expressive, and when is it not? Ms. Johnson. Senator, that question is being worked through with the courts right now. And it is an incredibly challenging question. I think one of the reasons that we have at the Knight Institute been looking to structural changes and structural regulation is because, as Ranking Member Schatz said earlier, you know, the law doesn't always keep up with technology. And as we try to figure out what is expressive and what is not, and we are waiting for the courts to help us with that, it is almost like a game of whack-a-mole, right. When--if we can actually look at some of the root causes of this, where we know these terrible things are happening and people are being drawn in because the platforms know so much about them, they kind of know what we are going to do before we know we are going to do it kind of thing, right? But if we have more research into what is happening, if researchers are protected, if we have specific data privacy protections, understanding how what is being collected and how it is being used, those are the issues that are going to help us, I think, try to solve this problem as quickly as possible. Senator Baldwin. It seems that part of the problem here is that every platform uses different algorithms to push content to users and have different protections in place for children on their platforms and have different policies on content moderation. I want to stick with you, Ms. Farid Johnson. How transparent are online platforms about their use of algorithms and granular moderation decisions? Would it be beneficial if Congress required increased disclosure and explanation for these practices and their uses? Ms. Johnson. So, I think there is a way to ensure additional transparency into the platform's practices. What we have said is that any kind of required transparency, any kind of disclosure should not be seen to be an undue burden on speech. But those can be crafted, certainly. I think in terms of whether we look at having additional understanding of what the platforms are doing would also benefit from allowing independent researchers in the public interest to conduct that research online so that they have access. And right now, they are really thwarted from doing so because the terms of service of these platforms are so onerous that they could be subject to civil and even criminal liability for their efforts. Senator Baldwin. Thank you. Thank you, Mr. Chairman. The Chairman. Thank you. Senator Curtis. STATEMENT OF HON. JOHN CURTIS, U.S. SENATOR FROM UTAH Senator Curtis. Thank you, and especially thank you, Mr. Chairman, for this hearing. It means a lot to me, and it means a lots to these parents. So, thank you. Mr. Bergman, you started your firm, I think, in reaction to what you were seeing out there from some of these parents that are here today. Section 230 was intended to protect us. And it is clear that other things get in the way. Maybe it is profits, whatever it is, from that. And you look for ways to hold companies accountable, and we have talked about these product features several times. They enlisted them, and I will come back to them in a minute. But I agree with that approach, and that is why I entered the Algorithm Accountability Act to clearly define that a platform who doesn't uphold a duty of care should not be protected by Section 230. So the question is, does Section 230 reform, does it come in conflict with the Second Amendment? And before you answer that, I will ask you to answer it, but I want put this in context with my second question. And Section 230, as I understand, was originally designed to protect platforms from the liability of a third party message or a third-party search, if that makes sense. And the best analogy I have been able to think of is the Post Office. So imagine for a moment that I mail a letter to you and the Post Office delivers that letter. We would never dream of holding the Post Office accountable for the content of that letter. But I think we do acknowledge that there are certain things that the Post Office would be justified in withholding that letter--child pornography or perhaps a bomb, right, or something that we give them that license in certain cases. And my experience after eight years here in Washington is that we all stop right there and say, how does the Post Office define whether or not they should deliver that letter. Let's suppose, hypothetically, they could read every letter. And so, we hear the Chairman say, and I totally agree with him, is we would want the Post Office to not stop any more letters than they have to. We would want as many letters to get through as possible. And we are in this--what I would call this foolish cycle of trying to define what should be delivered and what should not be delivered. And everybody has their own definition. You hear some, like the Chairman, say almost everything should be delivered. And then I hear other of my colleagues argue and say, well, why would you deliver that letter? But I want to take us past that point to the mailbox. Now, when that letter gets to the mailbox, imagine if the Post Office had opened that letter and read it, and then designed features for the homeowner of that letter, that they then sent not the letter to that person, but they sent it to 100 million people. That same letter, we said, oh, we like the content of that. We are going to send it to 100 million people. We like the contents because we think they will use the Post Office more if they read this letter. And then they design, and now we come into the features, push notifications, the automatic scroll, and all these things that we have been listed, that the Post Office is doing to get those 100 million people to read that letter. And later on top of that, the fact that we know that they are far more likely to distribute a letter that is sensational, that is--that does what we have talked about today. And so, you have defined that, and we are trying to define that as product liability. So, now let me come back to the question of, can we approach this without damaging the Second Amendment if we go to the end, right to the mailbox, where it is being opened up instead of this decision about, are they censoring? And we are talking about now the liability of their actions. And can you define the legal distinction between holding a platform liable for speech, which is like in the middle there, versus the platform liable for the way it distributes that speech? Mr. Bergman. Yes, Senator. And just to follow up on your analogy, it would be as though the letter had cocaine attached to it and the person would become addicted. But listen, in 1996 when Section 230 was enacted, the First Amendment had been around for 205 years and had protected the rights of free expression. The fact that Section 230 imposed a blanket liability--if that were to be modified, we would still have a robust body of law to protect free expression and the free exchange of ideas. And there is two elements to that. Number one, the question is, to what extent does AI constitute speech? And that is a very esoteric question. And the answer is sometimes yes, sometimes no. So, you know, algorithmic recommendations may or may not have a free speech component. The second question, though, Senator, and this is very important, is that even if something is speech, it doesn't mean that it is necessarily protected. The vile material that Selena Rodriguez received was speech, but it was online grooming, and it is clearly not protected. Libel isn't protected. And our courts have a very robust jurisprudence, and you know, there is nothing more, I think, impressive than Chief Justice Roberts' analysis in Snyder to analyze where a Tort action imposes speech and where it doesn't. So I think that if--if cases were allowed to go back, if cases that would be barred by Section 230 were allowed to proceed in the court system, the jurisprudence that we have on the First Amendment, I think, would draw that vital distinction between free speech and--protection. Senator Curtis. I agree with you. But let me point out, and a lot of my colleagues would point out, and you are lawyer, you would love this, right--there would be lots of lawsuits. So walk through with me the advantage of the legal system deciding this versus a Senator placed in a moment of time trying to say this is OK, but this isn't OK. Mr. Bergman. Well, the laws evolve over time, Senator. And you know, particularly defining what is or isn't protected speech is something that is conferred on the courts. I think that were Section 230 to--and I don't favor the repeal but allow cases such as these families to go forward, courts would still have the opportunity to apply time tested First Amendment jurisprudence to determine the extent to which those claims implicate protected speech and which cases don't. Senator Schatz. Versus us taking a point in time and then 30 years later decide if we got it right. Now--and I am out of time, but let me just conclude with this, a quote from Section 230, ``encourage the development of technologies which maximize user's control over what information is received by individuals families and schools.'' Are we meeting that standard of Section 230? Mr. Bergman. Unfortunately not, Senator. Senator Curtis. Yes, I wish you had more time. Thank you, Mr. Chairman. I yield. The Chairman. Thank you, Senator Curtis. Senator Lujan. STATEMENT OF HON. BEN RAY LUJAN, U.S. SENATOR FROM NEW MEXICO Senator Lujan. Thank you, Mr. Chairman. Mr. Carson, one question raised by the introduction of generative AI products is the extent to which companies should be held liable for illegal content generated by their products. As legal scholars point out under the current Section 230 regime, while platforms enjoy immunity, plaintiffs can still sue the original creator of illegal content. But AI generated search results that may cause harm could injure parties--could leave injured parties without recourse. Mr. Carson, what is your view on whether generative AI search engines fit into existing Section 230 framework? Mr. Carson. I don't believe that Genitive AI should be under Section 230 at all. Senator Lujan. Do you think platforms will need to verify that posts are authored by a human to benefit from Section 230 protection? Mr. Carson. No, if they are hosting someone else's use of AI, like say I make an AI image, and then post it on Facebook, that is clearly covered by Section 230 today. But the AI system itself, if they were to use--what ChatGPT creates for me personally, that is not covered by 230. That is their own speech and therefore not covered by 230. Senator Lujan. So Elon Musk using Grok to put a face on-- your face on a naked body is OK? Mr. Carson. No, definitely not OK in any way. And he probably has some 230 immunity for that. I wouldn't give him that immunity. I would make him subject, as Mr. Bergman said, to the common law on those kind of questions. But he would certainly claim 230 for his own posting, since he owns Grok today and X. So that would be an interesting legal question. Senator Lujan. I appreciate that. I hope it gets tested. Ms. Keller, in 2023, I authored an amicus brief in Moody v. Netchoice where I argued that social media companies have a First Amendment right to curate and moderate content on their platforms. Discussions regarding Section 230 often focus on the laws, liability, protections, but your testimony--in your testimony you write that, ``the choice to moderate is precisely what Section 230 encourages.'' And that, ``the law has produced a variety of platforms that have adopted diverse approaches to content moderation.'' Ms. Keller, you also filed an amicus brief in Moody v. Netchoice where you argue that Texas and Florida laws at issue burden the speech of the platforms. Can you explain to us the role Section 230 plays in the First Amendment rights of platforms to engage in content moderation? Ms. Keller. Certainly. So Section 230--platforms have rights to engage in content moderation, as the majority in the Moody case decided under the First Amendment. Section 230 basically gives an additional layer of procedural protection around that, so that if they are sued over these things, they can get the cases dismissed quickly, which if you are a startup, is incredibly important. These cases can get extremely expensive. Sorry, I feel like I am missing the other part of your question. It was about Gonzalez and---- Senator Lujan. I think you touched on it. I was getting more to the area that you included, the burden from the Texas and Florida law, which was in the same case. Ms. Keller. I see. And to be clear, the thrust of that brief was really about the rights of users being affected. About the State Governments coming along and saying, hey, if you want to use the Internet and look at--you know, hear the podcast that you want to hear, you have to navigate through all of this garbage that otherwise would have been weeded out by the platform. So, we were concerned with the rights of users, and we proposed that a less restrictive means than the state setting the rules would be middleware and interoperability, and something that allows users themselves to take control. Senator Lujan. Ms. Farid Johnson, a similar question to you. Do platforms have a First Amendment right to engage in content moderation? Ms. Johnson. Yes. Senator Lujan. In your testimony, you state that--again, Ms. Farid Johnson--your testimony states that Section 230 does not always provide liability protection. For example, certain platforms may be held liable for harms that arise from content neutral tools. Unfortunately, as Mr. Bergman points out in his testimony, social media companies will argue that every algorithmic design must be exempt from liability. And while there are certainly gray areas, it is hard for me to believe that a design feature solely in place to get and keep kids addicted to social media is immune from all liability. My question is, when does a design feature cross over from being an editorial decision immune from liability to a content neutral tool outside of Section 230 protections? Ms. Johnson. Thank you, Senator. That is a question that is actually making its way through the courts right now. We are seeing a number of cases that are asking the question about when a design is expressive, such as it would be protected under the First Amendment. And we just don't have the answers yet. I will say that in terms of algorithmic amplification, decisions that amplify are going to be quite difficult to distinguish from decisions to publish. So for example, if a newspaper were to put a particular headline, they are amplifying a particular piece of news because they are trying to get someone to read that news. So I think that it is important to look at the fact that there are going to be these decisions there that are going to seem editorial. And from our perspective, the way to address this is not necessarily to wait out the courts in terms of trying to figure this part out and not try to legislate on that, because you can't legislate around the First Amendment, but rather to say, we know what some of the structural issues are with respect to people having--the people's engagement online. And if we know that they are being targeted because of the information collected about them from these platforms, then the way to go about it is to address the data privacy issue rather than trying to circumvent the First Amendment. Senator Lujan. Let me ask you this, yes or no follow-up. Should social media companies provide the public with transparency regarding targeted algorithms and data collection? Ms. Johnson. Yes. I do believe that transparency provisions, if appropriately drafted, can certainly be Constitutional, yes. Senator Lujan. I appreciate that. And Mr. Chairman, I will submit this question to the record. Mr. Bergman, I was going to ask a question about through your litigation experience, how are social media companies weaponizing design decisions to keep kids addicted to social media? I will submit that into the record, and I look forward to hearing your response as well. Thank you for the time, Mr. Chair. Mr. Bergman. I will finish--thank you. The Chairman. Thank you. Senator Capito. STATEMENT OF HON. SHELLEY MOORE CAPITO, U.S. SENATOR FROM WEST VIRGINIA Senator Capito. [Technical problems]--thank you, Mr. Chairman, and thank all of you for your testimony today. Very interesting. I would like to address the parents of the three children. I think when we think of 30 years ago, when Section 230 was written and enacted, we could have never imagined that we would see three young people at such a young age take their own life through what we thought was a magnificent discovery in the 90s. And so, I want to thank them for what they are doing, because what you are doing is helping other parents as we see these kinds of things. You are so brave to do this. And we just had an instance of this. I live in West Virginia. An instance that you might have read about it. A 15-year-old boy took his own life very quickly in a sex-torsion case, and his parents have come forward. So, I know it can't be easy. And it leads me to my other thought. So we are sitting here. We have got Kids Online Safety. We had a hearing on the addictive nature of social media. We had four information sessions on what AI, and what does that mean, and how good is it, and how bad is it, but we haven't done anything. And so, I think part of our problem is if we legislate today in this area, we are not going to pick up what I learned day before yesterday, the new thing that is coming, which is super intelligent AI that is smarter than anybody in this room and all of us put together. And so, we are going to still have those gaps, I think, that are still going to be able to catch our young people and others into the entrapment to do damaging things to themselves. And so, I feel frustrated by that we can't figure out a better way to protect. And I understand the pushes and pulls of the First Amendment, but when I look at the families, I think we got to do better here. We can't just keep having hearings and wondering. And I know the Chairman--and there is lots of good legislation out here, but we really need your help as a professional panel to help us weave those challenges. And so, if you look at the companies, this just struck me. I thought if you looked at the company, why would they pursue this line of algorithm? Well, you know, it is advertising. It is money. And I started thinking about Chick-fil-A. You know what, Chick-fil-A has a belief system that they don't want to be open on Sundays. And that is to their own economic detriment, although I think they smoke all the other fast food places on 6 days than most of them do it on 7 days. So, it can be done. You can self-select. You can self- regulate and still be extremely successful in a competitive environment. So my question is, am I right to be concerned that whatever we could do today--I am just going to go down the panel--whatever we can do today, we are going to come back in 10 years and be obsolete? And so, I will start with the first-- Ms. Keller. Ms. Keller. The risk of obsolescence is indeed high. However, for AI, there is the backdrop of common law and tort law. Anything that is unprotected by Section 230, as some people have said, the output of generative AI would be, has this more flexible set of tools. Senator Capito. Yes, but aren't these companies using AI to generate their algorithm, or am I wrong there? Ms. Keller. I think they are using a mix of AI and other editorial inputs that the court told us were immunized in the Moody case. Senator Capito. Ms. Johnson. Ms. Johnson. I agree, the risk of obsolescence is high, which is why I think that looking at these foundational questions is critically important. If we have--even from now, if we were to establish an AI regulatory framework that was promoting independent research into the AI developers, if we understood more about what was happening on the platforms, if we limited the data they were collecting about us, those are the things I think could help move the needle in a way that would not run up against what happens next in terms of the next Big Tech thing. Senator Capito. Thank you. Mr. Bergman. Mr. Bergman. Yes, Senator. In MacPherson v. Buick, Justice Cardozo famously said the law changes--the laws of the stagecoach adapt to the age of the automobile. The common law of the states does provide the ability to adapt and apply traditional concepts of responsibility, of negligence, and products liability to ever-expanding technologies. So, the first and foremost thing that I believe Congress can do is allow the legal system to operate. And basically to use its economic function that Judge Posner said to internalize the cost of safety and impose on social media companies the same rules that every other company has. I think that would be a good start. Senator Capito. Thank you. And my former colleague, Mr. Carson. Mr. Carson. Senator, more than most on the panel, I think Section 230 was a mistake. And mostly, it is not because we don't need laws. It is both because, as Justice Thomas so well said, ``it is a get out of jail free card.'' And those are the kind of things you can't put in place before you actually understand how an industry is going to develop. So you need flexible laws. You shouldn't just freeze it in place at a time when the Internet was hardly developed. Senator Capito. Thank you. And again, thanks to the parents for being here. The Chairman. Thank you. Senator Rosen. STATEMENT OF HON. JACKY ROSEN, U.S. SENATOR FROM NEVADA Senator Rosen. Thank you, Chairman Cruz. Thank you to the witnesses for being here. And I just want to build on a point that Senator Capito made. And you know, I just don't believe we can allow just self-regulation and our common law to be our guide just because technology moves quickly. It doesn't mean that we here in Congress shouldn't legislate or regulate where necessary and adjust where needed. We can't cede our power because things move quickly. The world has always moved quickly to your point about the stagecoach to the car to, well, autonomous vehicles maybe soon. And so, I just want to make that comment for the record. But I want to talk about a risk of Section 230 protection for AI. Because last summer, I led a bipartisan task force--I led a letter with the bipartisan task force for combating Antisemitism to xAI, calling for accountability after their AI chatbot, Grok, went on multiple antisemitic tirades on X and spread conspiracy theories about the Holocaust. So, to reiterate what Senator Lujan asked Mr. Carson, should chatbots like Grok, when integrated into a social media platform, be protected by Section 230? And what are the risks if AI chatbots, which are sometimes seen as less biased and more accurate, are integrated on platforms and are protected under Section 230? Mr. Carson. I don't believe that you should have generative AI be considered as speech at all, and therefore it should not be protected by Section 230. In the case of Grok, in the example you gave, those companies should be liable under some kind of product liability theory, some kind of design defect theory for making products that are engaging in this kind of behavior. Senator Rosen. Thank you. I am going to move on and talk about speech then because Congress and the courts have determined that commercial speech by companies is not the same as free expression by an individual. Companies are liable for harm when they release unsafe products, like an unsafe car seat, an unsafe energy drink that causes heart attacks, and they can be held accountable for lying about their products. So Mr. Bergman, I am going to ask you this time, do you think a social media platform failing to enforce its own content moderation policies makes its products unsafe? And are these platforms marketing themselves as one thing by explaining their policies online, but then failing to create the online environment that reflects their policies? Mr. Bergman. I think that is very much the case. We just completed a trial in Los Angeles, and because we could get past Section 230, we were able to elicit and present documentary evidence that these companies intentionally have addicted children, knowing that children are being hurt because of it. This evidence directly contradicted the testimony of the executives before this very committee, and we think that that is a very important development and a reason why Section 230 should not preclude these cases from going forward and let the truth be heard. Senator Rosen. Thank you. And then we will move on to impact on other platforms, because some websites, like Wikipedia, Reddit, they rely on decentralized content moderation. It is partially why their communities are so vibrant and why people use their websites. So there is a concern that eliminating Section 230 could jeopardize their entire model. So Ms. Johnson, we will move down to you. How can Congress hold the largest social media companies accountable, while protecting novel content moderation approaches and smaller websites that operate differently from the big players? Ms. Johnson. Senator, you are exactly right that Section 230--when we say platforms, we often mean social media. But it really, it applies to quite a broad swath, which is why when the Knight Institute has considered this question, we have come up with three proposals looking at researcher access, data privacy, and interoperability as possible means of conditioning to Section 230 protection for the very largest platforms, and thereby allowing others to continue to thrive. Senator Rosen. Thank you. I am going to move back to you, Mr. Bergman, and ask you this. Some larger platforms would distinguish, again--many of which have designed their products to maximize engagement over all other metrics. Keep those eyeballs on. They have fired content moderation staff and are some cases no longer taking down content that violates, again, their own policies, and they still claim Section 230 is necessary. So should there be a different standard for our larger platforms? Mr. Bergman. I think everyone should have the duty of reasonable care. And I think Section 230 does provide important protections. And some should stay, but it should not be interpreted outside of what Congress intended when it enacted the statute in 2000--I am sorry, 1996. Senator Rosen. Thank you, I yield. The Chairman. Thank you. Senator Hickenlooper. STATEMENT OF HON. JOHN HICKENLOOPER, U.S. SENATOR FROM COLORADO Senator Hickenlooper. Yes. Thank you, Mr. Chair. And thank each of you for being here today and for all your work on this. This is obviously an issue of great complexity but of almost unimaginable importance as we really look at how deeply this damage goes. And I know this issue around broad immunity and 230, you guys are already debating pretty well. I am not going to opine on it. I am not sure Government is smart enough or can move fast enough to regulate AI at the speed it is going. And I don't see another way to deal with the situation without allowing or compelling the industry to regulate itself, which is really what 230 gets in the way of. In other words, if tort law can take actions against damages done to groups of people, history shows us that that is the way you create standards and you create the evolution of standards to meet the needs of that moment. In terms of my time with you now, I will give my own little speech. I want to talk a little bit about supply chain designations. And this is a little obscure, but I think no less important. Really, for the first time ever, the United States has designated an American company as a supply chain threat and blacklisted it from working with the U.S. Government. This is the same level that has been given to China's Huawei, Russian cybersecurity firms like Kaspersky. It is ironic to see that our Administration that fears censorship under Section 230, and claims to want a smaller Government, is now trying to weaponize Federal Government law to force a company to dismiss their own policies and prevent AI from being used for conducting mass surveillance against Americans, for lethal targeting that is done autonomously. So, Ms. Keller, why don't we start with you, but I would like to go down the list of just, what message does it send to a supply chain risk designation--what is that message to an American company--that is given to an American company--what is the message it sends to the rest of the innovation economy, let's put it that way? Ms. Keller. It sends a message that the Government is willing to bully and retaliate and make improper uses of laws to punish those who dissent, or disagree, or attempt to hold reasonable limits. Senator Hickenlooper. Thank you. Ms. Johnson. I think to the extent that this type of action is taken in retaliation for speech, it sounds an incredibly challenging--chilling message. It is very chilling. Mr. Bergman. It is really not in my ambient to opine on that. Mr. Carson. Senator, we have worked a lot on that issue. It is an indefensible decision that will chill the innovation economy and an attempt to murder one of the leading instruments of national power that our country has, which is an incredible AI lab at Anthropic. Senator Hickenlooper. Right. It is certainly something that is a place of power that we are giving away, in a way. Let me switch there. Ms. Johnson, thank you for being here and your advocacy of the First Amendment rights--your ability to defend the First Amendment. Some have argued that the way an AI model responds to a prompt from a user is a form of editorial discretion. In this case, that is certainly protected by the First Amendment. But if we imagine a scenario where Section 230 liability shield projections for AI outfits were removed, when we think about that, Congress, as we debate about how to establish AI accountability and an accountability framework, how could this be done without violating the First Amendment-- the protections on speech? Ms. Johnson. It is an incredibly challenging question, Senator, and it is something that is being worked through-- working its way through the courts now. I think I will respond in two ways. One is that if you are looking at Section 230 protections specifically, Section 230 insulates platforms from liability for publication of third party content. So, to the extent that AI output, is that of AI companies? It does not fall under that provision. In terms of crafting a durable piece of legislation that would help with AI accountability, I think what is important to start with are transparency provisions. If we had provisions that promoted and protected independent research into these AI developers to really understand how they operate, to understand what is going on behind the scenes, that would provide Americans with information that they can use in the public interest that could ultimately then lead to additional accountability. Senator Hickenlooper. Right. I agree. Thank you. I yield back the floor. Thank you. Senator Markey. STATEMENT OF HON. EDWARD MARKEY, U.S. SENATOR FROM MASSACHUSETTS Senator Markey. Thank you, Mr. Chairman. Thank you for having this hearing. And again, I just want to recognize the parents who are here, who have lost their children. It is almost unimaginable, the loss which you have suffered. And thank you for being here. Thank you for bearing witness to this mental health crisis in our country amongst teenagers and children. It is just very important that you are here. It is almost impossible to fully understand the grief you must feel right now, and it is my honor to stand with you in fighting to protect other families from the same thing that you have gone through, because those harms underscore the importance of the issues before us today. And I am glad that we are having this hearing, because it is obviously very controversial, and it just raises so many important issues of online safety, and privacy, and free speech. As the different views are being laid out here today, we can see that this is a subject that we should be talking about. So I would like to--Mr. Bergman, if you could. I know that your firm has done a lot of thinking about this issue and how to protect young people online while working within Section 230 in the lawsuits which you are bringing. Can you explain further to the Committee how your firm is bringing cases against the platforms and avoiding dismissal on Section 230 grounds? Mr. Bergman. Well, avoiding the dismissal sometimes, not other times. But yes---- [Inaudible.] Mr. Bergman.--avoiding dismissals sometimes, not other times. But we follow the theory enunciated by the Ninth Circuit in Barnes v. Yahoo, and then Lemon v. Snap, that we focus on the design, and that it is not the content. It doesn't matter what the algorithm--the algorithms don't care what they show the kids. Whatever addicts them to the platform. It could be moonbeams and rainbows, as long as a kid becomes addicted to this platform through operant conditioning. So we focus on the addictive design. We focus on infinite scrolls. We focus on the likes and the streaks features. We focus that these companies, as we have learned in this litigation, deliberately target kids, knowing that their brains are not fully developed and that they are very susceptible as adolescents to peer pressure. Senator Markey. So you are saying the conduit itself? Mr. Bergman. That is correct. The platform itself is dangerous. Senator Markey. It is dangerous, yes. And I do--I appreciate your approach to this work to protect kids online. Because ultimately, you know, we are going to have to try to deal with this issue. And this committee, with the Chairman and the Ranking Member, were very instrumental with Senator Cassidy and I in passing COPPA 2.0, Child Online Privacy Protection Act, through the Senate floor just two weeks ago, unanimously. Could you talk a little bit about how important COPPA 2.0-- do you agree it is important for us to pass the Child Online Privacy Protection Act in order to guarantee that they can't target kids with ads, that your parents can demand that everything be erased, that we raise it up to age 17 and under 17? Do you do agree that should become a law in our country, Ms. Keller? Ms. Keller. Can adults have that too, please? Senator Schatz. I am with you. And I actually passed that on the House floor in 1995. I am going to go take a note in the conference committee with the Senate, but I did pass it for adults as well. Ms. Keller. Thank you. Senator Markey. Across all platforms I got that passed 30 years ago. It was highly anticipatable. Yes, Ms. Johnson. Ms. Johnson. Yes. Yes, robust protection for children online in terms of private protections is critical--yes. Senator Markey. Thank you. Mr. Bergman. Absolutely, Senator. And the bipartisan leadership of this committee has been instrumental. This committee really has been the fulcrum of bringing these issues to the fore over the last five years under the bipartisan leadership. And on behalf of the families, you have really already saved a lot of lives. Senator Markey. Thank you. And then Mr. Carson. Mr. Carson. It would certainly be a great improvement. Congratulations to the Committee for their leadership on this. Senator Markey. And I am right now working on AI chatbots legislation. Does anyone want to talk about that and the importance for us to legislate in there? Mr. Bergman. Mr. Bergman. Our firm brought the first case involving AI chatbots, involving a 14-year-old boy who was goaded into suicide through an online chatbot. And we successfully overcame a First Amendment challenge and were able to move forward. We are bringing the first cases against OpenAI for the same thing. And again, this is basically a design flaw. We know that AI is here to stay, and it does a lot of good. But companies need to take proactive measures to think about safety. Senator Markey. These kids are becoming emotionally dependent upon AI chatbots. And so, we need to move in that area as well, passing legislation. Mr. Carson, would you like to add on to what Mr. Bergman just said? Mr. Carson. Yes, I think it is a critical issue. I think it comes to Congress making clear, and this addresses Senator Hickenlooper's issue as well, the outputs of ChatGPT, or Grok, or Gemini are not protected speech under the First Amendment at all and are not worthy of the protections of the First amendment as expressions of human creativity or human autonomy. And that yes, the rules around those chatbots are very important for children especially, but for the broader society. Senator Markey. Yes. As technology moves, we have to move as well. The technology is inanimate. It is only as good or bad as the human values that we instill into those inanimate objects. And we have to continue this conversation on chat-- this new area is absolutely something that we should be discussing and moving on in this committee as well. Thank you, Mr. Chairman, for having this hearing. The Chairman. Thank you. Senator Blackburn. STATEMENT OF HON. MARSHA BLACKBURN, U.S. SENATOR FROM TENNESSEE Senator Blackburn. Thank you, Mr. Chairman. And thank you to each of you for being here today. It has been a wonderful discussion, and we appreciate your insight. And to the parents, I want to say thank you for once again being here and for the advocacy that you bring. I appreciate that Senator Markey talked about COSA and Kids Online Safety, and Senator Cruz has been such an advocate for protecting children in the virtual space. And indeed I wish that our friends in the House were as committed to getting some of this legislation across the finish line as we are here in the Senate. Mr. Bergman, I want to come to you. As I have listened to the hearing today, thinking back through where we were in the mid-90s and the advent of Section 230, which seemed like a really great idea for something that was going to be an unknown, if you will, with the virtual space and giving companies a chance to get their sea legs under them, but what we have seen is massive abuse of Section 230. And the way that these social media platforms and Big Tech, as they have grown, they have become more given to making excuses for their actions and blaming it on Section 230, that it allows them to do this, that, and the other. As I have thought through this--and one of the reasons I have grown to be in support of sunsetting and removing 230 is because Big Tech has proven they are incapable of regulating or policing themselves. They will not do it. They are like an errant child who keeps pushing, and pushing, and pushing, and trying to move away any kind of responsibility, any discipline, and they fight it every single day. And we have seen it as we have worked with parents. We have seen it as we talk to pediatricians and principals who talk to us about behavioral issues in school that all find what happens online and the online platforms do nothing. So talk for me a little bit about Big Tech's refusal to take an action to protect, and the need that that puts on Congress to take an action to force them to protect? Mr. Bergman. Well, Senator, first of all, thank you from the bottom of my heart for your steadfast, emphaticable efforts on this issue. The kindness and the compassion and the commitment is an inspiration to all of us and thank you for that. We just finished a trial in which we saw--and we have now been able, because we got over Section 230 at least a little bit, to be able to see the internal documents from these companies. And we see that indeed there are people of conscience within these companies sounding the alarm bells. And time and time again, their calls go unheeded because any time a design change would impair profitability or engagement, they say no. How many times have the executives been excoriated before your committee and they don't change their behavior? How many times have they had bad press? The only thing that is going to change their behavior is when they have to bear the economic costs of their deliberate design decisions. As, you know, Richard Posner or Milton Friedman would say, you have to internalize the cost of safety. If they have to bear the cost of their dangerous platforms instead of these families, instead of clergymen, and policemen, and doctors, and psychologists, and insurance companies, then they will have the incentive to change their behavior. But right now, there is no way. If they actually had to bear cost through a civil lawsuit, their behavior would change. And I think unless--and through the imposition of civil liability, we can change their economic calculus. And we are not talking about, you know, imposing a special duty. We are just talking about imposing the same rules that every other company has. Every other company in America operates under a duty of reasonable care. We are asking the same thing for social media. Senator Blackburn. Well, and I think it is important to note that every industrial sector in the United States has safety standards. They have that duty of care. The only one that does not is the virtual space. There are no safety standards. That is why COSA is a safety by design. It is a product safety bill to protect children in the virtual space. And it is unseemly to me that we have this growing industrial sector, and they have zero safety standards. And honestly, they don't give a rip and flip. When you have got somebody like Mark Zuckerberg who says each kid is worth $270 to him, that is one of the most offensive statements I think I have ever heard come out of the mouth of a U.S. corporate CEO, to put a value on the head of every child that is using their product. It is just unseemly. We are hard at work, and I have just released the Trump America AI Act framework. The President asked me to take a shot at drafting this as just a guideline as to where we start on the discussion about a framework for AI. My hope is we are going to move forward on this more quickly than we did other components of the virtual space. And I would like for you to respond for just a minute about why you think it is important, Mr. Bergman, to have an AI framework as we begin to move forward with more AI concepts moving into commercialization. Mr. Bergman. Well, because we continue to see families that have buried children because AI chatbots encourage suicide. One would have thought after two and a half years, I could never be shocked. But when I saw what Sewell Setzer was provided, encouraged to kill himself, when I saw that Zane Shamblin was given a how- to manual--we have to do something, Senator. And your leadership is such an inspiration, and I am just so grateful. On behalf of all the people I represent, but also as a father and a grandfather, thank you. Senator Blackburn. Well, we appreciate you all. I want to thank each of you for being here today and for the spirited debate that you have brought to the issue. I will remind you that members of the Committee have until March 25 to submit their questions. You all will have until April 8 to respond in writing to those questions that will be coming to you. And with that, it concludes our hearing. Committee adjourned. [Whereupon, at 11:58 a.m., the hearing was adjourned.] A P P E N D I X Prepared Statement of Hon. Ron Wyden, U.S. Senator from Oregon Thank you, Chairman Cruz and Ranking Member Cantwell, for inviting me to submit a statement for the record for this important hearing on Section 230. For the past three decades, Section 230 has allowed critics, organizers and activists to use the Internet to make their voices heard. Online communities large and small, from knitting message boards to firearm enthusiast forums to neighborhood ICE Watch Facebook groups, are all only possible as a result of Section 230. Section 230 is best understood as the embodiment of a simple principle: speakers are responsible for their own speech. Absent such a rule, the incentives to host speech, especially controversial speech, disappear. No competent business would let users post freely on its websites if any single post could land it in court. If 230 were repealed, only the deepest-pocketed online platforms in the world could survive the avalanche of frivolous litigation that would follow. And on those platforms, only anodyne or low-value, likely AI-generated, content would pervade. However, because of 230, digital library platforms can host comments about books being read in their communities; individuals can buy domains and host online forums for their niche special interests; individuals can sell homemade goods through online marketplaces; and people with interesting ideas about how to remake social media can raise money to fund projects like Bluesky and Mastodon. What is more, 230s text explicitly protects users of platforms. Section 230 protects users of services like e-mail from facing lawsuits for forwarding article links. It protects Reddit moderators and Wikipedia editors from getting sued by members of their online communities or the subjects of their articles, respectively. And it protects anyone who `retruths,' `reskeets,' or otherwise reposts content on a platform. The ability to rapidly reshare information online is only possible because of 230. Critics of 230 often misunderstand that while the statute provides a great deal of immunity, it does not create a `get out of jail free' card to online services. 230 only protects companies when a lawsuit seeks to treat a company as the speaker of the post they find offensive or harmful. Courts can and have held companies liable for their own speech and business practices. For example, Amazon has tried and failed to use 230 to avoid lawsuits about dangerous batteries it helped sell. Meta also tried and failed to use 230 to dodge responsibility for helping place discriminatory ads. This logic extends to AI. Chatbots like Claude and ChatGPT are services that create content. While users might request a chatbot's output, the bot's response is the company's own speech. When a company creates or materially contributes to illegal content, the plain text of 230 says the company is responsible for it. When claims go to trial, 230 is no barrier to holding AI companies accountable. Section 230 is not perfect. There may be reasonable reforms worth pursuing that do not pressure services into censorship or invasive monitoring. However, changing the rules of the road for online speech would increase the power of Big Tech and billionaires who want to crowd out competition and censor critics. A far better path to a healthier online ecosystem is to focus on a few long overdue policy priorities. We must pass a baseline privacy law that would make it harder for platforms to target users, including children, with harmful content. And we must reform our antitrust laws to make sure that smaller platforms and services have a fighting chance to give users more choices on where to consume content and host their speech. Implementing these two reforms, and taking a close look at proposals to foster more transparency and interoperability, would drastically improve the online experience without endangering online speech. Instead of starting with the question ``How should Congress change Section 230?'', I'd urge lawmakers to ask, ``How can the Senate make the Internet work better for all users?'' This body has plenty of work to do on that front. ______ Prepared Statement of Brandon Arnold, Executive Vice President, National Taxpayers Union Section 230 Is Essential for Preserving a Thriving Internet As the ``Voice of America's Taxpayer,'' National Taxpayers Union (NTU) is a strong supporter of Section 230 of the Communications Act of 1996, which, for 30 years, has helped hundreds of millions of Americans amplify their voices on the internet. Section 230 has long been a lightning rod for criticism. While by no means perfect, it provides an important liability shield that protects online platforms of all sizes--not just ``Big Tech,'' as its detractors often claim. Section 230 has been called ``the Twenty-Six Words That Created the Internet,'' because, broadly speaking, it insulates platforms acting in ``good faith'' from undue litigation. The need for such a protection is clear. The Internet has created an unbelievable amount of opportunities for education, entrepreneurialism, entertainment, and other endeavors. It has, unfortunately, also created many dangers and risks, especially to young people. It has become a forum for heated debate over politics, religion, sports, and pretty much any conceivable topic. As these debates devolve, they have too often given rise to harassment, bullying, and even physical threats. For platforms, this has created tremendous challenges. Policing online behaviors, taking down inappropriate comments or posts, and responding to phenomena like ``review bombing'' have become critical-- and costly--functions for platforms. These functions expose companies to criticism. For instance, taking down a politically charged post could invite accusations of censorship, while allowing unsavory comments to remain on a message board could suggest that a company endorses that viewpoint. Some have suggested that repealing Section 230 would solve the problem. In actuality, it would make the situation substantially worse. Repealing Section 230 would completely alter the manner in which online platforms monitor and respond to reviews, comments, and posts. This would harm social media sites like Nextdoor, X, and Truth Social; review sites like Yelp and Rotten Tomatoes; video game platforms such as Steam and Roblox; and many, many more. To avoid litigation, platforms would likely respond by aggressively censoring a much larger quantity of user input. This could reduce the amount of information available to a consumer shopping for a new refrigerator, prevent a reader from posting comments about an op-ed he just read, or stop an aspiring musician from sharing her new song. A vastly more restrictive Internet would undermine creativity, stifle innovation, and harm public discourse. Additionally, Section 230 preempts many state laws, so its repeal could create a patchwork of state statutes governing liability related to online content. This would create confusion, complexity, and higher costs for all platforms, but would be especially burdensome on smaller companies that lack the resources to comply with dozens of regulations and an untold number of lawsuits. This could result in higher levels of consolidation of online companies and higher costs for end-users. Section 230 is essential for preserving a thriving Internet and this hearing provides an excellent opportunity to highlight that fact. It is incredibly important for members of this Committee to utilize this opportunity to share their views on Section 230 with the American people. At the same time, it is even more important for the hundreds of millions of Americans who are not members of this Committee to be able to continue to share their views on the topic of their choice in the Internet ecosystem. The latter would not be possible without the prudent protections established by Section 230. ______ Prepared Statement of Laura Chadwick, President and CEO, Travel Technology Association On behalf of the Travel Technology Association, I submit the following written statement for consideration ahead of the ``Liability or Deniability? Platform Power as Section 230 Turns 30'' hearing. The Travel Technology Association (``Travel Tech'') represents travel technology innovators ranging from dynamic startups, small and midsize businesses, to leading Online Travel Agencies, metasearch engines, short-term rental platforms, Global Distribution Systems, and Travel Management Companies. Section 230 of the Communications Decency Act plays a critical role in enabling modern online travel platforms to function safely and effectively by taking proactive steps to protect consumers. It allows platforms to host user-generated content--such as reviews, ratings, and photos--without facing liability for content posted by third parties, while also empowering them to manage and remove material that risks consumer safety. These protections make it possible for millions of travelers to see information shared by other travelers when planning a trip. Reviews, photos, and listings on online travel platforms help consumers compare options, evaluate quality, and make informed decisions about where to stay, how to travel, and what experiences to book. Together, these tools form a foundational part of today's travel marketplace. Because of Section 230, travel platforms are encouraged to take proactive steps to reduce the risk of harm to consumers. They can remove fake listings, potentially misleading reviews, and inappropriate content that could expose travelers to potential increased risk. They can display real-time user-generated information that helps travelers assess the quality and reliability of travel services. And they can ban or block users who engage in scams, harassment, or other behavior that undermines consumer trust. The importance of this information to consumers is clear. Surveys consistently show that 90 percent of travelers rely on information posted on platforms, such as online reviews, before making a booking. Nearly all travelers--approximately 95 percent--read reviews during the planning process, and many spend weeks researching options before finalizing travel decisions. Travel Tech members helped pioneer the use of user-generated content in the travel industry in the early 2000s, fundamentally changing how travelers research and book trips. Community-driven reviews and listings created a more transparent and competitive marketplace, where quality and service are visible to consumers and where small businesses can reach a global audience. As Congress considers Section 230, it is important to recognize the role these protections play in supporting consumer safety and transparency in travel and how, broad or sweeping changes could have unintended consequences. Weakening Section 230 protections may hamper platforms' ability to invest in proactive efforts to remove fraud, scams, bad actors, and misleading content. It may also force platforms to restrict or remove user-generated content in a more sweeping way, leading to fewer reviews, less transparency, and fewer tools for travelers to evaluate safety and quality before they book. In short, Section 230 helps create an environment where platforms can responsibly manage the information travelers depend on. Preserving that balance is essential to maintaining consumer trust, protecting the travel experience, and supporting a competitive travel marketplace. Thank you for your consideration. Laura Chadwick, President and CEO, The Travel Technology Association. ______ TechNet Washington, DC, March 17, 2026 Hon. Ted Cruz, Chairman, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Hon. Maria Cantwell, Ranking Member, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Dear Chairman Cruz and Ranking Member Cantwell: In advance of the Senate Commerce Committee's hearing titled ``Liability or Deniability? Platform Power as Section 230 Turns 30,'' I am writing to share TechNet's perspective on the continued importance of intermediary liability protections and the principles that should guide any reforms to Section 230 of the Communications Decency Act of 1996. TechNet is the national, bipartisan network of technology CEOs and senior executives that promotes the growth of the innovation economy by advocating a targeted policy agenda at the Federal and 50-state level. TechNet's diverse membership includes over 100 dynamic American businesses ranging from startups to the most iconic companies on the planet and represents over five million employees and countless customers in the fields of information technology, artificial intelligence, e-commerce, the sharing and gig economies, advanced energy, transportation, cybersecurity, venture capital, and finance. As Section 230 marks its 30th anniversary, TechNet urges the Committee to reflect on what these protections have made possible. Section 230 has been a cornerstone of the modern internet, enabling American innovation, supporting millions of jobs, and allowing businesses of all sizes to connect with customers, share ideas, and grow online. It has helped fuel the economic growth and global competitiveness that define the U.S. technology sector. These protections do not provide ``total immunity'' from wrongdoing. Rather, they ensure that bad actors, not the platforms they misuse, are held accountable for their own conduct, while preserving the legal certainty that has allowed decades of innovation to flourish. As the Committee examines potential reforms, TechNet urges members to be guided by the following principles: First, the First Amendment cannot exist in the 21st century without protections for the intermediaries that provide opportunities for user- generated speech. Making interactive computer services liable for content created by their users would force providers to censor broadly to protect themselves, reducing access to information, fundamentally changing how people communicate online, and making it more difficult to protect users, including children and vulnerable communities. This outcome would be bad for users, bad for innovation, and inconsistent with our Nation's commitment to free expression. Second, any reforms to Section 230 must not create uncertainty that disproportionately harms entrepreneurs and makes it harder for startups and emerging companies to compete. The large platforms that critics most often target have the legal and financial resources to absorb sweeping liability changes. It is the small businesses, startups, and new entrants that cannot, and it is those innovators who would be most harmed by repeal or poorly designed reform. Policymakers should hold bad actors accountable for their own actions rather than shifting liability to the platforms that host them. Third, algorithms are not publishing decisions. Automated systems that organize and surface content, whether recommending a product, a video, or a route, do not constitute an endorsement of that content. Treating algorithmic tools as editorial judgments subject to publisher liability would threaten the core functionality of search engines, social media platforms, online marketplaces, and countless other services that consumers rely on every day. Fourth, any reforms should support reasonable processes and systems for managing illegal content based on industry best practices, rather than imposing one-size-fits-all mandates that disadvantage smaller platforms and new entrants. Liability protections allow interactive computer services to set their own rules for content moderation in ways that best serve their users, and that flexibility is a feature, not a flaw. TechNet urges Congress to reject efforts to repeal Section 230 without a clear, workable alternative that protects consumers, preserves free speech, and ensures the United States remains the global leader in innovation. We stand ready to work with the Committee toward targeted reforms that hold bad actors accountable, protect vulnerable users, and preserve the legal certainty that allows new companies, technologies, and ideas to thrive. Thank you for the opportunity to submit these views for the record. Please do not hesitate to reach out if we can be a resource or if you have any questions. Sincerely, Linda Moore, President and CEO. ______ LGBT Tech Tuesday, March 17, 2026 Hon. Ted Cruz, Majority Leader, United States Senate. Hon. Maria Cantwell, Minority Leader, United States Senate. Dear Chairman Cruz, Ranking Member Cantwell, and Members of the Senate Committee on Commerce, Science & Transportation: On behalf of LGBT Tech, thank you for the opportunity to submit this statement for the record in support of Section 230 of the Communications Decency Act. LGBT Tech is a national nonprofit working at the intersection of the LGBTQ+ community and technology. Our work focuses on digital equity, privacy, online safety, and ensuring LGBTQ+ people can access the tools and spaces they need to thrive. Section 230 is a core protection that makes online speech and community spaces possible, and is essential to protecting LGBTQ+ people's ability to speak and connect online. We recognize that online harms are real and any policy response must be carefully tailored to those harms. However, eliminating or destabilizing the current legal framework, including Section 230, that allows online services to host and moderate lawful user content is not the tailored response users need. It is a structural change with predictable harms that will be borne first and hardest by communities that already face censorship, exclusion, and disproportionate barriers to information. That is why LGBT Tech is especially concerned that the debate is no longer limited to targeted reform proposals that ``update'' Internet law but is increasingly focused on proposals to sunset or repeal Section 230 altogether. For LGBTQ+ people, that would have immediate and foreseeable effects. At a time when attacks on LGBTQ+ people are intensifying in schools, legislatures, courts, and other physical spaces, digital platforms increasingly function as lifelines. They are places where people seek identity-affirming information, connect with peers, find crisis support, access health and safety resources, and participate in public discourse. If Section 230 were repealed or dramatically weakened, many platforms, especially smaller services, nonprofits, forums, and community-based spaces would likely respond by narrowing what users can post, restricting discussions on topics deemed controversial or high risk, or suppress entire categories of content altogether--including LGBTQ+ issues--rather than risk being sued and incur expensive litigation. That chilling effect would not fall evenly across communities. In the current political climate, LGBTQ+ identity and even the mere existence of affirming resources are frequently characterized as inappropriate or harmful, especially when minors are involved. If liability risk increases dramatically as a result of the dilution or elimination of Section 230, LGBTQ+ content and resources would be among the primary categories pushed toward over-removal and exclusion from platforms. Our concerns are also reflected in the community we serve. LGBT Tech's 2025 polling of LGBTQ+ people in the United States found that 78 percent were concerned about the removal or restriction of LGBTQ+ content by social media platforms, rising to 97 percent among transgender respondents. 76 percent of respondents (and 94 percent of transgender respondents) reported concern about losing access to online LGBTQ+ community spaces because of platform policy decisions. Weakening Section 230 would significantly increase the risk of censorship and isolation that our community members already fear. For LGBTQ+ people, especially those living in hostile or isolated environments, access to digital spaces is essential to safety and community. As Congress considers the future of Section 230, we urge the Committee to recognize that weakening or repealing it will not simply increase accountability. It will predictably narrow access to some of the last remaining spaces where LGBTQ+ people can safely speak, connect, and be seen. Respectfully, Shae Gardner, Director of Policy and Research. ______ Center for Countering Digital Hate March 17, 2026 Hon. Ted Cruz, Chairman, Committee on Commerce, Science, and Transportation, United States Senate. Hon. Maria Cantwell, Ranking Member, Committee on Commerce, Science, and Transportation, United States Senate. Re: Committee Hearing on Section 230 Reform Dear Chairman Ted Cruz and Ranking Member Maria Cantwell: I write to you as the U.S. Director of Public Affairs for the Center for Countering Digital Hate (CCDH), a watchdog organization that researches the societal harms associated with social media and artificial intelligence platforms. I was a child, growing up in Central Minnesota, when Section 230 was signed into law. I witnessed the Internet's transformation from early message boards and instant messaging to today's large-scale social media platforms. What began as relatively simple online communities has evolved into sophisticated systems that track users' behavior, relationships, and preferences to maximize user engagement and advertising revenue. Alongside the rest of the nation, we are closely following ongoing litigation against Meta in New Mexico and Los Angeles. In New Mexico, prosecutors allege Meta violated consumer protection laws and failed to address addiction and child-exploitation risks on its platforms. In Los Angeles, a bellwether youth-addiction case has proceeded on product- design theories, and Mark Zuckerberg testified before a jury in February 2026. Section 230 issues may remain central on appeal, particularly where courts are asked to distinguish between liability for third-party speech and liability for product design or platform conduct. While Congress does not have authority over the confines of the First Amendment, it does very much have authority over Section 230. As we detail below, policymakers need only examine court rulings from the past three decades to see the growing shortcomings of Section 230 as the United States' primary legal framework for governing the modern Internet. Courts, even within the same circuit, are increasingly divided over how the law should apply. Justice Elena Kagan suggested during Supreme Court oral arguments that Congress may be better suited than the courts to address questions about Section 230s scope. Big Tech Needs to Come to the Table to Negotiate To be clear, we are not calling for a repeal of Section 230, but rather a thoughtful congressional review of potential reforms that better address Americans' expectations for justice and accountability. In that spirit, we support the Sunset Section 230 Act (Graham/Durbin) as a mechanism to force tech companies to engage Congress and stakeholders in good-faith negotiations towards meaningful reform. Members of this very Committee have put forward legislative solutions that deserve thoughtful consideration. The need for legislation such as the Sunset Section 230 Act reflects the technology industry's limited engagement with Congress on meaningful reform. Major companies and their trade associations-- including NetChoice and the Computer & Communications Industry Association--have often resisted proposals to revisit the scope of Section 230. Meanwhile, Americans continue to face serious risks online, including scams targeting seniors and safety threats affecting children and teens. The Broken ``Good Samaritan'' Compromise Section 230 of the Communications Decency Act was originally enacted with a straightforward goal: to encourage Internet companies to moderate harmful content on their platforms. In 1996, Representatives Christopher Cox and Ron Wyden also sought to address a controversial New York state court ruling that threatened the ability of online platforms to host user-generated content in good faith. Early message board operators justifiably needed safeguards to remove posts that were irrelevant or objectionable (for example, posts unrelated to dog care on a dog care forum) and to prevent being held liable for user- generated content, such as defamatory restaurant reviews, that platform operators had no reasonable way of knowing were unlawful or protected speech. In exchange for these ``Good Samaritan'' actions, Congress granted the then-nascent Internet industry a degree of immunity from liability for user-generated content. The highly sophisticated social media business model and technological capabilities that exist today were not envisioned by the original proponents of Section 230. Over time, however, the Good Samaritan compromise has been undermined by judicial interpretations that now permit the world's largest and most profitable corporations to harm its own users with impunity. Section 230 Raises Significant Federalism Concerns The current debate surrounding artificial intelligence regulation illustrates another important concern. Americans overwhelmingly oppose Federal policies that preempt states from protecting their residents from the harms of the tech industry. According to a poll conducted by the Institute for Family Studies, three-to-one of polled Americans opposed the proposed inclusion of AI preemption in the NDAA. Yet Section 230 represents the earliest and most sweeping example of Federal preemption benefitting Meta, X, and other tech companies who are dominating in both the social media and AI sectors. In our large and politically diverse nation, states are often the laboratories of democracy. Federal lawmakers should remember that the Framers of the Constitution intended for most governing authority to remain with the states. Section 230 has often prevented states from pursuing remedies for harms including scams, fraud, deepfake pornography, sexual exploitation, harms to children and teens, defamation, and other harms facilitated through social media platforms. Two dozen state Attorneys General filed an amicus brief in the Gonzalez v Google case before the Supreme Court, arguing that Section 230 violated sacrosanct principles of federalism and was preventing states from allocating ``losses for Internet-related wrongs'' amongst private parties.\1\ State Attorneys General know from personal experience that social media companies have the most power and resources to mitigate harms in the first place. It is time for Congress to address the flaws of Section 230 once and for all. --------------------------------------------------------------------------- \1\ Brief for the States of Tennessee et al., as Amici Curiae in Support of Petitioners at 9, Gonzalez v. Google LLC, 143 S. Ct. 762 (2023) (no. 21-1333). --------------------------------------------------------------------------- Restore Judicial Authority to Enjoin Illegal Activity In Hassell v Bird (2018), the California Supreme Court held that Section 230 prevents courts from ordering platforms to remove illegal content--in that case, defamatory material. While it is important not to place platforms in the impossible role of judge and jury on the First Amendment, it is quite another matter to allow a platform or any company to ignore a lawful court order. It is essential for Americans to have certain protections online to protect their reputations, careers, finances, and personal safety. As the dissent correctly notes, the order sought to enforce a defamation judgment against the speaker and protect a victim of unlawful defamation, not to punish Yelp for publishing third-party content. Society is not served by allowing defamatory material to continue to spread. This holding has particularly acute consequences for the Jewish community, which has long been the target of defamatory conspiracies and incitement. Anti-Jewish hate frequently manifests through false allegations (e.g., modern forms of blood libel) against prominent Jewish individuals and institutions, as well as coordinated campaigns to spread disinformation that portrays Jewish people as malicious or criminal actors. When courts are unable to compel the removal of content that has already been adjudicated as defamatory, these narratives are allowed to persist, amplify, and reach wider audiences, increasing the risk of real-world harassment and violence. Leading voices in the space of combatting antisemitism have all warned that online platforms are amplifying anti-Jewish hate at scale. The continued circulation of defamatory content, despite any legal findings of its illegality, undermines both accountability and safety, leaving Jewish individuals especially vulnerable. Attacks on synagogues in the United States have been preceded or accompanied by the spread of conspiratorial and defamatory narratives online, underscoring how unchecked digital amplification of hate can translate into offline threats and physical danger for Jewish communities. Congress should clarify that Section 230 does not prevent courts from enforcing lawful judgments and ordering the removal of adjudicated illegal content. Hold Big Tech Accountable for Algorithm-Fueled Proliferation of Scams, Fraud, Misleading Advertisements, and other Harms What began as a modest protection meant to address the ``moderator's dilemma'' has evolved into a doctrine that courts often interpret broadly in favor of sweeping immunity. For example, in Force v. Facebook, the Second Circuit held that Facebook's use of algorithms to match posts and information to users based on their interests constituted a traditional ``publisher'' function, allowing the company to claim Section 230 immunity. Similarly, in Calise v. Meta Platforms, the Ninth Circuit granted immunity to Meta for the proliferation of user-targeted fraudulent third-party advertisements placed on its platform, reasoning that if a duty owed by the platform would require it to moderate content, Section 230 immunity applies.\2\ --------------------------------------------------------------------------- \2\ 103 F.4th 732 (9th Cir. 2024). Note, the contract claims surrounding Meta's terms of service were not barred by Section 230. --------------------------------------------------------------------------- These legal interpretations risk insulating platforms from responsibility even when their systems actively contribute to the spread of something as harmful as scams, fraud, and misleading advertisements. Certain courts have gotten it right. In Anderson v. TikTok, the Third Circuit held that algorithmic recommendations that curate and amplify content can constitute a platform's own conduct to be ineligible for the protections of Section 230. At a minimum, Congress should consider a ``deliberate indifference to harm'' standard that addresses situations in which platforms possess knowledge that harmful content is causing real-world harm but fail to take reasonable steps to mitigate it. Mitigate the Harms of Design Features Congress should clarify that Section 230s liability shield applies to third-party, user-generated speech, not to a platform's own conduct, product design and interface choices. The Federal circuits have not reached a consistent standard on this issue. The Ninth Circuit has taken steps to avoid creating what it called a ``lawless no-man's-land on the Internet,'' however, the Second Circuit, by comparison, has favored broader Section 230 immunity regarding design features immunity.\3\ --------------------------------------------------------------------------- \3\ HomeAway.com v. City of Santa Monica, 918 F.3d at 683 (9th Cir. 2018). --------------------------------------------------------------------------- In Fair Housing Council of San Fernando Valley v. Roommates.com, the Ninth Circuit held that platforms are not entitled to Section 230 immunity when their design choices materially contribute to the unlawfulness of third-party content.\4\ Similarly, in Lemmon v. Snap, the Ninth Circuit allowed claims to proceed against Snapchat's speed filter, concluding that the case concerned the negligent design of a product with a potential defect rather than liability for user speech.\5\ --------------------------------------------------------------------------- \4\ 521 F.3d 1157 (9th Cir. 2008). \5\ 995 F.3d 1085, 1091-94 (9th Circ. 2021). --------------------------------------------------------------------------- Clarifying that Section 230 does not shield harmful platform design decisions would preserve protections for user speech while ensuring companies remain accountable for the systems they build. Conclusion It is well past time for Americans to receive a legislative process that fully considers both the benefits and the harms produced by Section 230. Congress should do better in reassessing how this framework from three decades ago fails to function in today's vastly different technological environment. Section 230 was not covered by the nightly news or debated in the newspapers at the time. There is little legislative record to suggest that the members of this Committee in 1996 contemplated a future in which children would spend hours each day on platforms designed to track their online activity, addict them for higher engagement quotas, and push harmful self-harm content to them within minutes of creating an account. The legislative history is noticeably thin for what is now called the ``Twenty-Six Words that Created the Internet.'' We at CCDH respectfully urge this Committee to continue its examination of Section 230, seek input from a broad range of stakeholders, and consider meaningful legislative reforms. Sincerely, Sarah O'Quinn U.S. Director of Public Affairs. ______ Consumer Technology Association March 17, 2026 Hon. Ted Cruz, Chairman, Committee on Commerce, Science, and Transportation, United States Senate, Washington, DC. Hon. Maria Cantwell, Ranking Member, Committee on Commerce, Science, and Transportation, United States Senate, Washington, DC. Dear Chairman Cruz, Ranking Member Cantwell, and Members of the Committee: This year marks the 30th anniversary of the Communications Decency Act, enacted on February 8, 1996. Section 230 of that Act laid down a simple principle: liability for speech rests with the speaker, not platforms hosting third-party content. Section 230 supercharged American tech leadership by creating clear rules and legal protections allowing innovators to build. Over the past 30 years, those 26 words helped power the modern Internet and cemented American global leadership in the digital economy. Since then, American Internet companies have generated trillions of dollars in market value and created millions of jobs. Today, American tech companies account for roughly one-third of the S&P 500s total market value. The United States also produces more venture-backed startups than any other country and attracts the largest share of global venture capital investment. Seven of the ten largest tech firms globally are American. We have the world's most dynamic digital economy, and Section 230 made that possible. Section 230 provided legal certainty allowing innovators to invest and take risks. Without it, every website that hosts a comment, review, rating, forum, or marketplace listing would face constant lawsuits over user speech, creating a crippling litigation tax on the internet. Certainty matters most for startups and early-stage companies. Large companies can hire teams of lawyers. Startups and smaller businesses-- like 80 percent of the Consumer Technology Association's more than 1200 member companies--cannot. Section 230 is not a subsidy for large platforms. It is the legal infrastructure that allows thousands of startups, creators, and small businesses to launch platforms, grow communities, and attract investment. Weakening Section 230 protections would not discipline large platforms. It would entrench them, making it nearly impossible for the next generation of competitors to thrive. Many of the companies and technologies that debut each year at CES, our annual globally tech event, would not exist without Section 230. CES features over 4000 exhibitors introducing new products, platforms, and services to the global market, showing vibrant competition in the technology sector. Section 230 helped create the legal environment that allows those innovations to emerge and scale. Section 230 also helped enable the greatest expansion of speech in human history. Before the internet, reaching a large audience required owning a television or radio station, a newspaper, or another printed publication. Section 230 provided the legal certainty allowing platforms to host user-generated content at scale. Millions of Americans can now publish their ideas, grow audiences, launch businesses, and participate in public debate online. Creators, educators, entrepreneurs, artists, journalists, and small businesses all benefit from this open and competitive ecosystem. Section 230 also protects platforms ability to moderate and remove objectionable content in good faith. That flexibility allows different communities to develop different norms. LinkedIn operates differently from X. Yelp operates differently from Truth Social. That diversity reflects marketplace competition and innovation as companies tailor their content moderation practices to their users' needs and preferences. Section 230 does not protect illegal conduct. If something is illegal offline, it is illegal online. Fraud, trafficking, threats, and exploitation remain fully subject to criminal law enforcement. Major platforms invest billions of dollars each year in detecting and removing illegal content, and Section 230 enables that proactive work by ensuring that good faith content moderation does not become a liability trap. As with every significant new technology, the expansion of online participation and speech also creates new challenges. Protecting children online is, can and should be a priority. Thankfully, practical solutions exist. For example, the Digital Safety Shield proposed by NetChoice outlines steps that strengthen online safety while preserving free expression and innovation. Finally, Section 230 has renewed importance in the age of artificial intelligence (AI). Modern AI systems rely on vast amounts of third-party information and operate on top of the open Internet ecosystem that Section 230 helped make possible. AI systems summarize, organize, recommend, and interact with information created by millions of users across the internet. Without the protections Section 230 provides for hosting and organizing user-generated content at scale, the data and interaction layers that power modern AI would face enormous litigation risk. Weakening that foundation would slow American AI innovation precisely when global competition is accelerating. The United States won the first era of the Internet because Congress created clear rules for innovation. At a moment of intense global competition in AI and advanced technology, Congress should preserve the legal framework that made that leadership possible. We look forward to working with the Committee to ensure the United States remains the world's leading home for innovation, entrepreneurship, and free expression online. Sincerely, Kinsey Fabrizio, President, Consumer Technology Association. Michael Petricone, Senior Vice President, Government Affairs, Consumer Technology Association. ______ The Sidney Group March 18, 2026 Hon. Ted Cruz, Chairman, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Hon. Maria Cantwell, Ranking Member, Senate Committee on Commerce, Science, and Transportation, Washington, DC. RE: Statement for the Record for the Hearing, ``Liability or Deniability? Platform Power as Section 230 Turns 30'' Dear Chairman Cruz, and Ranking Member Cantwell: On behalf of The Sidney Group,\1\ I respectfully submit this statement for the record for the Committee's March 18, 2026 hearing, ``Liability or Deniability? Platform Power as Section 230 Turns 30.'' I submit this statement to flag one underexamined issue in the current Section 230 debate. Section 230 was built for a world in which platforms hosted third-party speech. Today, digital systems increasingly rank, shape, and generate the content at issue. That shift raises liability questions Congress should take seriously before litigation forces the issue first. --------------------------------------------------------------------------- \1\ The Sidney Group is a Washington, D.C.-based law firm with a focus on public policy advisory, technology policy, artificial intelligence governance, digital regulation, and regulatory risk in Federal and state policy environments. --------------------------------------------------------------------------- I do not believe Section 230 should be repealed. Blunt repeal would do more harm than good. It would create legal uncertainty, chill innovation, and discourage lawful online speech and moderation. But if Congress continues to debate Section 230 only as a content-moderation statute, it risks missing the next pressure point: how intermediary liability frameworks apply when systems increasingly help produce the content at issue. A Changing Technological Environment Platforms today do more than host content. They rely on recommendation systems, ranking systems, automated moderation tools, and increasingly generative AI systems that shape what users see and, in some cases, generate the content itself.\2\ The core issue is no longer only whether platforms host harmful content, but whether existing liability frameworks can still map cleanly onto systems that increasingly help produce it. --------------------------------------------------------------------------- \2\ See Gonzalez v. Google LLC, 598 U.S. (2023). --------------------------------------------------------------------------- Section 230, however, was drafted with the assumption that the relevant speech at issue would be provided by another ``information content provider,'' that is, a source distinct from the platform itself.\3\ --------------------------------------------------------------------------- \3\ 47 U.S.C. Sec. 230(f)(3). --------------------------------------------------------------------------- Generative artificial intelligence systems complicate that structure. Large language models and related systems can generate text, images, audio, and other forms of content directly in response to prompts or system inputs.\4\ Once integrated into digital services, they may participate in producing the very content users encounter online. --------------------------------------------------------------------------- \4\ National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework 1.0 (Jan. 2023); Exec. Order No. 14110, Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence Sec. 3 (Oct. 30, 2023). --------------------------------------------------------------------------- Emerging Liability Questions As generative AI systems become more widely deployed, courts are beginning to confront new questions about liability frameworks originally designed for human-generated content. At bottom, these cases raise a common question: when the system helps generate the content at issue, existing intermediary doctrines may no longer fit as neatly as they once did.\5\ --------------------------------------------------------------------------- \5\ Gonzalez v. Google LLC, 598 U.S. (2023) (considering whether claims arising from platform recommendation algorithms are barred by Section 230); Mark A. Lemley & Bryan Casey, Remedies for Robots, 86 U. Chi. L. Rev. 1311 (2019). --------------------------------------------------------------------------- That is not a theoretical problem anymore. Cases involving recommendation algorithms have already tested the boundaries of Section 230 where platform systems play an active role in shaping information flows.\6\ More recently, lawsuits involving generative AI systems themselves have begun to emerge. For example, a Florida mother sued Character.AI and Google after her 14-year-old son died by suicide in 2024, alleging that a chatbot portraying itself as a therapist and romantic partner worsened his mental state; a Federal judge later allowed key claims in the case to proceed.\7\ Another wrongful-death lawsuit alleges that Google's Gemini chatbot contributed to the suicide of a 36-year-old Florida man after the system allegedly fostered an emotional attachment and encouraged self-harm.\8\ --------------------------------------------------------------------------- \6\ Force v. Facebook, Inc., 934 F.3d 53 (2d Cir. 2019) (holding that Section 230 barred claims based on Facebook's use of algorithms to recommend and amplify third-party content); Gonzalez v. Google LLC, 598 U.S. (2023) (addressing whether Section 230 applies to claims arising from algorithmic recommendation of third-party content). \7\ Garcia v. Character Technologies, Inc., complaint filed Oct. 2024 (M.D. Fla.); Katie Paul, Google, AI firm must face lawsuit filed by a mother over suicide of son, U.S. court says, Reuters (May 21, 2025). \8\ Lawsuit says Google's Gemini AI chatbot drove man to suicide, Reuters (Mar. 4, 2026) (reporting wrongful-death allegations that an AI chatbot fostered emotional dependency and encouraged self-harm). --------------------------------------------------------------------------- Courts are being forced toward this question whether Congress is ready for it or not. Not all of these cases arise directly under Section 230. But they illustrate the kinds of AI-enabled harms, and the kinds of content- generation questions, that increasingly place pressure on legal frameworks originally designed to address harms arising from online speech and platform-mediated content. Related litigation over AI training data and outputs also shows courts beginning to examine how existing legal frameworks apply to systems capable of producing new content from large training datasets.\9\ --------------------------------------------------------------------------- \9\ Andersen v. Stability AI Ltd., No. 3:23-cv-00201 (N.D. Cal.) (alleging that generative AI models were trained on copyrighted works and produce outputs derived from protected material); Getty Images (US), Inc. v. Stability AI, Inc., No. 1:23-cv-00135 (D. Del.) (asserting claims arising from the use of copyrighted images in AI training and output generation); see also Getty Images (US), Inc. v. Stability AI Ltd. (U.K. High Court) (raising similar claims concerning training data and AI-generated outputs). --------------------------------------------------------------------------- In that sense, the next pressure on Section 230 may come less from abstract calls for repeal than from the practical difficulty of applying a hosting-era statute to systems that rank, shape, and generate content. What Congress Should Focus On Congress need not resolve that question fully in this hearing. But it should recognize that the Section 230 debate is no longer only about hosting, moderation, and repeal. It is increasingly about whether a liability framework built for platforms hosting third-party speech can still map cleanly onto systems that rank, shape, and generate content. If Congress waits, the issue may be shaped first by fragmented litigation and diverging judicial opinions rather than by deliberate legislative judgment. Because this use case is still new, Congress has an opportunity to examine it early and with a scalpel. In practical terms, Congress may wish to examine three related issues: how intermediary liability doctrines apply when automated systems generate content; how responsibility should be allocated across developers, platforms, deployers, and enterprise users; and how emerging AI systems are placing pressure on legal frameworks designed for an earlier Internet architecture. Section 230 was written for a world where platforms hosted speech. The next legal questions arise in a world where systems increasingly help produce it. Thank you for the opportunity to submit this statement for the record. Sincerely, Arthur D. Sidney. CEO and Principal, The Sidney Group. ______ Response to Written Questions Submitted by Hon. Eric Schmitt to Daphne Keller On March 18, 2026, during the Senate Commerce Committee's hearing on Section 230, you and I had an exchange concerning Missouri v. Biden, the Stanford Internet Observatory, and the Election Integrity Partnership. In that exchange, you stated that my description of Stanford's role was ``not consistent with [your] understanding,'' said that Stanford actors were ``exercising their First Amendment rights,'' later acknowledged, ``I wasn't there. I can't tell you how those conversations went,'' and added that you ``can't imagine them doing the government's bidding.'' When I referenced Missouri v. Biden, you also asked: ``The one that you lost?'' Since that hearing, the United States District Court for the Western District of Louisiana has entered a consent decree in Missouri v. Biden imposing restrictions on certain Federal agencies' ability to coerce or pressure social media companies to suppress protected speech, and the government agreed that Plaintiffs are prevailing parties. In light of that development, please answer the following questions. Question 1. During the hearing, after I referenced Missouri v. Biden, you responded: ``The one that you lost?'' I replied at the time that the case had been remanded on standing grounds and that ``we didn't lose the case.'' a. Do you retract your statement suggesting that Missouri v. Biden was a case Plaintiffs ``lost''? Answer. I do not retract that statement. In Murthy v. Missouri, the Supreme Court held that the plaintiffs, despite extensive discovery, failed to marshal enough evidence to establish standing to seek a preliminary injunction. As Justice Barrett wrote, plaintiffs' claims ``cannot rest on `mere allegations,' but must instead point to factual evidence.'' The Murthy plaintiffs did not meet this standard, the Court concluded, because they did not show ``any concrete link between their injuries and the defendants' conduct[.]'' The Court also noted that many of the district court's factual findings ``appear to be clearly erroneous.'' I believe this constitutes a loss before the Supreme Court.\1\ --------------------------------------------------------------------------- \1\ As stated in my testimony, I consider Murthy to be a problematic ruling that may harm future plaintiffs with legitimate claims against the government. Some commentators, such as the Wall Street Journal editorial board, have criticized the ruling more harshly, but recognized the bottom line: ``President Biden won[.]'' A Supreme Court License for Social-Media Censorship, Wall Street Journal (June 26, 2024), https://www.wsj.com/opinion/murthy-v-missouri-supreme- court-censorship-social-media-facebook-covid-samuel-alito-amy-coney- barrett-1ea4feda; see also Amy Howe, Justices side with Biden over government's influence on social media content moderation, SCOTUSBlog (June 26, 2024), https://www.scotus blog.com/2024/06/justices-side- with-biden-over-governments-influence-on-social-media-content- moderation/. --------------------------------------------------------------------------- In subsequent proceedings, as I understand, this evidentiary shortcoming was never redressed. Instead, the plaintiffs and current Administration reached and received district court approval of a settlement by consent decree. b. Do you acknowledge that the case has now resulted in a binding Federal court order restricting certain Federal agencies' conduct and in the government's agreement that Plaintiffs are prevailing parties? Answer. After my March 18 testimony, the plaintiffs and current Trump Administration defendants in Murthy ``reached an agreement resolving all claims'' and moved to enter a consent decree.\2\ The parties attached a draft order reflecting their agreement.\3\ In it, the government ``agree[d] not to contest that Plaintiffs are prevailing parties for purposes of 28 U.S.C. Sec. 2412(d)[,]'' the Federal statute governing attorneys' fees. The settlement specified that ``except as stated explicitly in the text of the Agreement, it shall not be construed as evidence or as an admission regarding any issues of law or fact, or regarding the truth or validity of any allegation or claim raised in this action or in any other action.'' --------------------------------------------------------------------------- \2\ Consent Motion to Enter Consent Decree (March 24, 2026), https://www.courtlistener.com/ docket/63290154/477/missouri-v-biden/. \3\ Exhibit A to Consent Motion to Enter Consent Decree (March 24, 2026), https://www.court listener.com/docket/63290154/477/1/missouri-v-biden/. --------------------------------------------------------------------------- The district court approved the settlement, adopting the parties' own negotiated language as a consent decree and order.\4\ As described by one expert observer, the order essentially re-stated the constitutional standards described by the Supreme Court in NRA v. Vullo, a successful jawboning claim heard the same day as Murthy.\5\ As its resolution of Murthy makes clear, the Court saw no sufficient evidence that defendants in Murthy had violated those standards in the first place. --------------------------------------------------------------------------- \4\ Order on Motion for Miscellaneous Relief (March 25, 2026), https://www.courtlistener.com/ docket/63290154/478/missouri-v-biden/. \5\ Morgan Wilsmann, Public Knowledge, Murthy v. Missouri Settlement Does Nothing to Combat Illegal Jawboning (March 30, 2026), https://publicknowledge.org/murthy-v-missouri-settle ment-does-nothing-to-combat-illegal-jawboning/. --------------------------------------------------------------------------- This negotiated settlement document is not a court holding. It does not reflect a judicial determination that any party prevailed. At the time of settlement, there was little meaningful adversity between the two parties, given the change in administration since the case was filed. Indeed, in light of a Trump Administration Executive Order titled ``Restoring Freedom of Speech and Ending Federal Censorship,'' the 5th Circuit had remanded Murthy's consolidated Kennedy v. Biden claim ``for the district court to consider mootness'' given the risk that the ongoing case failed ``the Article III requirement that there be a live case or controversy.'' \6\ --------------------------------------------------------------------------- \6\ Unpublished Opinion on Petition for Rehearing en Banc (5th Circ. January 30, 2025), https://www.courtlistener.com/docket/63290154/ 429/missouri-v-biden/ Question 2. During the hearing, when I described Stanford's role in the relevant censorship ecosystem, you stated: ``That version of the facts is not consistent with my understanding. I don't think they were --------------------------------------------------------------------------- doing the Biden administration's bidding.'' a. What specific facts did you rely upon in concluding that Stanford was not ``doing the Biden administration's bidding''? b. Please identify whether those facts were based on personal knowledge, secondhand representations, public reporting, institutional affiliation, or some other source. Answer. My understanding of these facts is grounded in publicly available information and my personal knowledge of the people involved in the Stanford Internet Observatory (SIO). Stanford's brief to the Supreme Court in Murthy describes SIO's operations in detail, as well as identifying numerous factual errors in the district court's findings. I cannot personally attest to that brief's level of detail, but I am confident of my assertion about the core question regarding the independence of people involved. As I stated in testimony, ``I can't imagine them doing the government's bidding.'' Question 3. In the same exchange, after expressing confidence that Stanford was not doing the government's bidding, you also stated: ``I wasn't there. I can't tell you how those conversations went,'' and ``I can't imagine them doing the government's bidding.'' a. If you were ``not there'' and ``can't tell how those conversations went,'' on what basis were you able to express confidence that Stanford was not acting in support of government censorship objectives? b. Do you wish to revise any portion of your March 18 testimony in light of your admitted lack of firsthand knowledge? c. Looking back, do you agree that your testimony conveyed greater certainty than your personal knowledge supported? Answer. I stand by my testimony and do not wish to revise it. As explained, my confidence comes from knowing the people at SIO. As I also testified, I fully support their First Amendment right to say what they think, including in speech expressed to the government. This is a broad principle, and applies to my colleagues' and others' First Amendment protected speech regardless of my personal agreement with any particular positions they may advance. The Fifth Circuit recognized this First Amendment concern in vacating aspects of the initial district court injunction that barred officials from ``collaborating, coordinating, partnering, switchboarding, and/or jointly working with the Election Integrity Partnership, the Virality Project, the Stanford Internet Observatory, or any like project or group[,]'' and in declining plaintiffs' subsequent request to reinstate this portion of the injunction.\7\ Such an injunction, the Fifth Circuit noted, ``may implicate private, third- party actors that are not parties in this case and that may be entitled to their own First Amendment protections.'' \8\ --------------------------------------------------------------------------- \7\ Plaintiffs-Appellees' Petition for Panel Rehearing, Case 23- 30445 (September 22, 2023). \8\ Missouri v. Biden, 83 F. 4th 350, 396 (5th Cir. 2023). Question 4. During the hearing, you rejected my description of Stanford's role as ``not consistent with [your] understanding,'' while also making clear that you were ``not really'' involved in the Stanford Internet Observatory's participation in the Election Integrity Partnership and that you were ``not there'' for the conversations at --------------------------------------------------------------------------- issue. a. Before offering that testimony, did you review any internal Stanford Internet Observatory or Election Integrity Partnership communications concerning requests, referrals, escalations, or coordination with government officials? b. If not, why did you present your understanding with confidence? c. If so, please identify the categories of materials you reviewed. Answer. My preparation for this testimony was focused on Section 230 and on what I consider to be very serious threats to the First Amendment posed by the current administration. It did not include review of materials of the sort described in the question. My confidence stems from my familiarity with the character of individuals involved in SIO, as well as public information including the record and resolution of the Murthy case. Question 5. Because your testimony addressed not only legal doctrine but also the factual character of Stanford's role in the censorship debates surrounding the 2020 election and COVID-era speech, the Committee would benefit from understanding whether your views have changed. a. Does the March 24 consent decree in Missouri v. Biden cause you to reassess the constitutional risks associated with coordinated relationships among Federal agencies, outside researchers, universities, and platforms in the content-moderation context? b. If your answer is yes, please explain how your views have changed. c. If your answer is no, please explain why the entry of a binding Federal consent decree does not alter your assessment. Answer. My views have not changed. The current Administration's decision to settle the case through consent decree does not involve a court ruling on the merits. ______ Response to Written Questions Submitted by Hon. Maria Cantwell to Daphne Keller Claims of Anti-Conservative Censorship: Some of my colleagues frequently claim that tech companies systematically censor conservative viewpoints, pointing to moderation decisions around January 6th or COVID-19. But treating content moderation purely as a partisan grievance ignores a fundamental reality of the internet--platforms must have the ability to make moderation decisions to create the kinds of communities their users actually want. We are seeing the demand for this right now with the rapid growth of heavily curated spaces like Bluesky. That is not to say platforms always get it right. In fact, they frequently make terrible decisions. We've seen catastrophic failures of moderation--like Grok and X allowing non-consensual intimate imagery to proliferate, or TikTok's algorithm feeding self-harm videos to children. Moderating content at scale is incredibly complex and often deeply flawed. But forcing platforms into a ``must-carry'' model to appease political complaints doesn't fix those severe harms--it makes them much worse. Ms. Keller, if Congress forced platforms to carry all speech regardless of their own community standards, what happens to their ability to address harmful content targeting children or that promotes self-harm or eating disorders? Answer. If Congress forced platforms to carry literally all speech, they would be unable to remove even worst-of-the-worst unlawful content such as child sexual abuse material. I believe that most proposals would not go so far, but would instead leave platforms able to enforce legal standards without being able to enforce their own discretionary standards. In that case, under the recent FSC v. Paxton ruling, platforms could conceivably be compelled to remove or age-gate pornography. But FSC would not support a legal mandate for age-gating non-pornographic material, including promotion of self-harm or eating disorders. Accordingly, we should assume that such content would be readily available to both adults and minors, regardless of their own preferences, under such a law. To predict what content would confront users in this situation, we must understand the wide array of ``lawful but awful'' speech protected by the First Amendment.\1\ The Supreme Court has held that the First Amendment protects lies,\2\ hate speech,\3\ and hurtful messages directed at grieving families which I believe would deeply offend most Americans regardless of political alignment.\4\ Lower court cases tell us that the First Amendment also protects diet advice\5\ that led to the death of a plaintiff's family member, and mushroom-collection\6\ advice that led plaintiffs to consume poisonous mushrooms that nearly killed them. All of this content--along with spam, scams, pro-terrorism content, barely-legal threats or incitement to violence, and more-- would remain on platforms under a legal standard of this sort. Content that people want to see would be harder to find, or visible only after navigating this tide of garbage. Few constituents would, I believe, thank their elected representatives for such a change. --------------------------------------------------------------------------- \1\ Daphne Keller, Lawful but Awful? Control over Legal Speech by Platforms, Governments, and Internet Users, U. Chi. L. Rev. Online (June 28, 2022), https://lawreviewblog.uchicago.edu/ 2022/06/28/keller-control-over-speech [https://perma.cc/5ZYJ-YLHC]. \2\ U.S. v. Alvarez, 132 S.Ct. 2537 (2012). \3\ Matal v. Tam, 137 S.Ct. 1744 (2017). \4\ Snyder v. Phelps, 131 S.Ct. 1207 (2011) (protestors at a soldier's funeral with signs saying ``God Hates You'' and ``Thank God for Dead Soldiers'' along with homophobic slurs). \5\ Smith v. Linn, 386 Pa. Superior Ct. 392 (Pa. 1989). \6\ Winter v. G.P. Putnam's Sons, 938 F.2d 1033 (9th Cir. 1991). --------------------------------------------------------------------------- Separately, if I may respond to one point from the Senator's introductory observations about Bluesky, I would like to call attention to what I believe to be that platform's single most important attribute. While the current primary interface may be heavily curated, Bluesky is revolutionary in that it can be flexibly curated. As an example, one Bluesky variant is Blacksky, a project focused on African American users, which enforces its own independent rules.\7\ I discuss the promise of this decentralized and interoperable approach in responses to Senator Schatz's questions, below. --------------------------------------------------------------------------- \7\ Blacksky Terms of Use, https://blacksky.com/blacksky-terms-of- use/. President Trump has sued multiple news organizations for coverage he claims is biased against him. If Section 230 were repealed, would those kinds of lawsuits have an even greater chilling effect on what platforms and media organizations are willing to publish? Answer. Platforms are now a news source for a majority of Americans, as well as a major forum for discussion of current events.\8\ Both users' access to news and their ability to discuss and debate its significance would be deeply threatened in a world without Section 230. Platforms would face legal and economic pressure from both the government and private claimants, such as business or political leaders discussed in news coverage. If those individuals claimed that coverage was false and defamatory, for example, platforms would have little ability or incentive to judge whether those claims are legitimate. President Trump has successfully targeted news organizations both in his capacity as a private plaintiff and through Federal agencies. --------------------------------------------------------------------------- \8\ Nieman Lab, For the first time, social media overtakes TV as Americans' top news source (June 2025), https://www.niemanlab.org/2025/ 06/for-the-first-time-social-media-overtakes -tv-as-americans-top-news-source/. --------------------------------------------------------------------------- Section 230 enables platforms to resist this pressure. Tellingly, when CBS's fear of FCC retaliation led to the elimination of a Colbert Report interview with a political candidate, the show's online version--on YouTube, a Section-230-immunized platform--was able to air the footage.\9\ Major platforms' courage in the face of such state pressure has clearly faltered in some cases, as I discussed in my testimony. But those concessions would surely be far greater without Section 230s immunities. Section 230 also ensures that, even if major platforms with more state entanglements yield, smaller and braver platforms can preserve access to news reporting. --------------------------------------------------------------------------- \9\ BBC, Stephen Colbert says CBS spiked interview with Democrat over FCC fears (February 2026), https://www.bbc.com/news/articles/ cze0dk3yd5eo. --------------------------------------------------------------------------- Carr Pressure and Jawboning: Earlier this month, FCC Chairman Brendan Carr once again threatened to revoke broadcast licenses for press coverage that the President does not like. This time it was because the press is reporting the facts about the Iran war. What's unsettling is that we have seen this type of behavior from Chairman Carr before--whether it's political commentary on late-night or reporting about ICE raids in a community. I agree with Chair Cruz that this type of coercion is ``dangerous as hell.'' It is a threat to free speech and the freedom of the press. Do you agree that a sitting government official using the threat of license revocation to suppress news reporting he and the President don't like represents one of the most serious threats to the First Amendment and free speech that we face today? Answer. I agree. Chairman Carr's outrageous behavior threatens a key institution of our democracy, and is one of the most serious assaults on the First Amendment today. Even before the most recent episode, eleven former FCC officials, including President Reagan's Chairman of the FCC, voiced their alarm that ``the current FCC Chairman has asserted the power to do precisely what the Supreme Court and Congress have forbidden'' by acting as a ``self-appointed, free-roving arbiter of truth in journalism.'' \10\ The FCC's current abuses of authority are unprecedented and profoundly dangerous. --------------------------------------------------------------------------- \10\ Petition for Special Relief In the Matter of Repeal of the News Distortion Policy (November 2025), https://protectdemocracy.org/ wp-content/uploads/2025/11/News-Distortion-Petition-for -Special-Relief-1.pdf. --------------------------------------------------------------------------- Section 230 Liability Protection: Section 230 was adopted thirty years ago on a simple idea: platforms hosting user speech should not be treated as publishers of every post, and those that moderate harmful material should not be punished for doing so. That framework has been critical for startups, small businesses, and online communities that depend on the ability to host user content without being sued over every post. But thirty years of experience have also shown that the largest platforms used these protections very differently from how Congress intended. They built business models around algorithmic amplification, addictive design, and pervasive data collection. They hid behind Section 230 while their products harmed children and disrupted elections. As we consider Section 230 at thirty, I want to understand what has worked well and what hasn't. Specifically, what these protections have meant for free speech and for the smaller platforms that depend on them the most. But also how we hold tech platforms accountable for their harms. Ms. Keller, how has Section 230 enabled online discussions about contentious but constitutionally protected topics, such as abortion, LGBTQ rights, or speech critical of the government? Answer. I am old enough to remember a pre-Internet media environment, in which a small handful of gatekeepers permitted only a trickle of speech to be widely shared. On broadcast TV, for example, experiences and perspectives of minority or marginalized groups were almost nowhere to be found. That profoundly limited account of the world existed for two major reasons. First, bandwidth was limited, and networks could make the most money with ``lowest common denominator'' content--material that would be appealing to the most people, but almost definitionally disserved marginalized groups. Second, that content needed to be reviewed by lawyers. Networks risked liability for any mistakes, and were incentivized to play it safe. Offering bland sitcom pabulum was cheaper and simpler than giving the microphone to speakers from diverse groups and perspectives, or carrying out investigative journalism about topics like police mistreatment, bias in hiring and workplace policies, or government misconduct. The Internet changed that, with Section 230 playing a crucial role. Section 230 ensures that gay and transgender teens in small towns can connect with broader communities and seek important information online. As one acute example, it protects platforms from the campaign against transgender identity that today's FTC is actively pursuing.\11\ It permits frightened and vulnerable pregnant teenagers to learn about options for abortions or even basic medical care outside their own states.\12\ It provides the most fundamental infrastructure for political organizing and dissent, which may rely on mundane 230- protected services like Dropbox, Microsoft 365, Google Spreadsheets, online e-mail, calendaring tools, chat groups, and more. Organizers of the Tesla protests explained this in a letter begging Congress to ``leave Section 230 alone.'' \13\ As they put it, without Section 230, the social media platforms we used to share information would have deplatformed us, for fear of being sued by Elon Musk or his supporters. And the web hosting companies and tools we relied on to allow anyone anywhere to create their own protest would have taken us offline. Being able to organize online together is a source of our strength and resilience. Don't destroy it. --------------------------------------------------------------------------- \11\ Nate Raymond, Reuters, Medical groups sue over U.S. FTC launching gender-affirming care probe (February 2026), https:// www.reuters.com/legal/government/pediatricians-group-sues- over-us-ftc-launching-gender-affirming-care-probe-2026-02-17/; Punchbowl News, As chair, FTC commissioner touts he'd pull back on AI and fight trans care (December 2024) (linking to leaked document, https://punchbowl.news/wp-content/uploads/FTC-Commissioner-Andrew-N- Fergu son-Overview.pdf, stating intention to ``[f]ight back against the trans agenda''), https://punchbowl.news/article/tech/ftc-commissioner- artificial-intelligence-trans-care/. \12\ Lizzie Presser and Kavitha Surana, ProPublica, A Pregnant Teenager Died After Trying to Get Care in Three Visits to Texas Emergency Rooms (November 2024), https://www.pro publica.org/article/nevaeh-crain-death-texas-abortion-ban-emtala. \13\ Sheridan et al, Tesla Takedown Organizers Letter: ``Democrats, don't kill the tools we need to organize!'' (April 2025), https:// www.fightforthefuture.org/news/2025-04-15-tesla-takedown -organizers-letter-democrats-dont-kill-the-tools-we-need-to-organize/. Do you agree that large platforms need to be held accountable for their actions and the harms they are causing to society--particularly harms to children and data privacy? Answer. Let me start with data privacy. I absolutely agree that platforms should be accountable for their use of personal data. Our country is--as Senator Cantwell knows far better than I do--long overdue for Federal privacy legislation. To the extent that children's harms genuinely sound in privacy and are caused by misuse of personal data, it is appropriate for Federal law to provide them with even broader protection. An update to COPPA that achieves this goal would be commendable. The cause of children's privacy is not advanced, however, by laws that cause platforms to implement age assurance or age verification measures. Such laws may cause platforms or vendors to collect anything from biometric data to behavioral profiles to copies of government IDs--from both children and adults. That information is at risk of data breaches, which have already happened and will surely continue.\14\ Perhaps more alarmingly, they are exposed to disclosure demands from governments and law enforcement. --------------------------------------------------------------------------- \14\ Josh Taylor, The Guardian, Proof-of-age ID leaked in Discord data breach (October 2025), https://www.theguardian.com/games/2025/oct/ 07/discord-data-breach-proof-of-age-id-leaked. --------------------------------------------------------------------------- This privacy and surveillance risk is the reason that Supreme Court case law prior to 2025's politically divided 6:3 ruling in FSC v. Paxton rejected such measures. Research shows that Internet users significantly change their behavior when they fear state surveillance.\15\ American journalists concerned about surveillance in the wake of the Snowden revelations, for example, engaged in ``levels of self-censorship . . . approaching the levels reported by writers living in authoritarian'' regimes.\16\ Google search queries dipped, including for purely personal queries about topics like health.\17\ Platform laws that impose this burden and risk on children and adults are extremely difficult to justify, both ethically and constitutionally. --------------------------------------------------------------------------- \15\ Alex Marthews and Catherine Tucker, Government Surveillance and Internet Search Behavior, (2014) (examining changes in Google search queries), https://papers.ssrn.com/sol3/papers .cfm?abstract_id=2412564; Jonathan W. Penney, Chilling Effects: Online Surveillance and Wikipedia Use, 31 BTLJ 117 (2016) (examining changes in Wikipedia page views), https:// btlj.org/data/articles2016/vol31/31_1/ 0117_0182_Penney_ChillingEffects_WEB.pdf. \16\ Summer Lopez, PEN America, Global Chilling: The Impact of Mass Surveillance on International Writers (2015), https://pen.org/report/ global-chilling/. \17\ Marthews and Tucker, supra note 15. --------------------------------------------------------------------------- ______ Response to Written Questions Submitted by Hon. Amy Klobuchar to Daphne Keller Areas of Agreement on Section 230: Platform design decisions like autoplay for videos, infinite scroll, and constant push notifications may not be protected by Section 230 and could be areas ripe for reform. Do you believe these are areas that Congress could effectively create safeguards without implicating speech issues? Answer. These are difficult and important questions. I am honestly uncertain of the answers, and glad that careful jurists in U.S. courts are currently reviewing evidence and arguments to resolve them. I will start with the question about speech. Regulation of such features would not be without First Amendment implications. For users, it would mean that state intervention made their speech less likely to reach an audience. As I examined in a law review article, the Supreme Court has repeatedly held that ``burdens'' on distribution of protected speech trigger the same First Amendment review as outright bans.\18\ Such measures would also restrict platforms' own speech, by preventing them from sharing precisely the sort of curated feeds at issue in Moody. --------------------------------------------------------------------------- \18\ Daphne Keller, Amplification and Its Discontents, 1 J. Free Speech L. 227 (2021). --------------------------------------------------------------------------- A court might conclude that restrictions on autoplay, infinite scroll, or notifications are not content-based, and hence are subject only to comparatively lenient intermediate scrutiny. On the other hand, both the intention and the consequence of such laws would likely be to burden certain speech based on its content. Lawmakers would not be inspired to act if social media feeds primarily featured dry academic lectures or CSPAN footage. Platforms would likely argue that in their case, the restriction is both content-based and unfairly targeted to disfavored speakers--hence meriting strict scrutiny. The other question is about the effectiveness of such measures. Would significant harm truly be avoided just by eliminating autoplay, infinite scroll, and notifications? It is unclear how real-world platform usage would change with these design alterations. Scientists remain divided about the causal link between social media use and mental health harms to teenagers.\19\ Addiction specialists question whether excessive social media use truly fits the ``addiction'' rubric.\20\ Some advocates for the rights and wellbeing of children champion legal restrictions on social media,\21\ others oppose them.\22\ Imposing restrictions solely for children would introduce privacy and surveillance risks, with the serious downsides discussed above. I am not an empiricist, and can't claim authority to unravel this jumble of potential positives and negatives. But I do think that Congress should not act without solid grounds to believe that doing so would actually achieve lawmakers' goals, while avoiding disproportionate harms to speech. Congress would also be wise to prioritize measures, like basic privacy protections, that can protect users from harm without engaging these fraught and difficult First Amendment questions. --------------------------------------------------------------------------- \19\ Mike Masnick, TechDirt, Two Major Studies, 125,000 Kids: The Social Media Panic Doesn't Hold Up, https://www.techdirt.com/2026/01/ 21/two-major-studies-125000-kids-the-social-me dia-panic-doesnt-hold-up/ (January 2026) (reporting on studies in JAMA Pediatrics and The Journal of Public Health). \20\ Ian Anderson and Wendy Wood, Nature, Overestimates of social media addiction are common but costly (November 2025), https:// www.nature.com/articles/s41598-025-27053-2. \21\ Common Sense Media, In Victory for Kids and Families, Newsom Signs Online Safety Protections into Law (2025), https:// www.commonsensemedia.org/press-releases/in-victory-for-kids-and- families-newsom-signs-online-safety-protections-into-law. \22\ Student Press Law Center, Amicus brief in NetChoice v. Bonta (N.D. Cal. 2023), https://splc.org/2023/05/amicus-brief-in-netchoice-v- bonta-n-d-calif-2023/. --------------------------------------------------------------------------- ______ Response to Written Questions Submitted by Hon. Brian Schatz to Daphne Keller Section 230s Impact on Censorship: Some experts have raised concerns that repealing Section 230 could result in platforms over-moderating content, which may disproportionately impact content deemed ``controversial.'' What content would be most vulnerable to over-moderation and censorship? How might these online harms translate to the real world? What has history shown us? Answer. Section 230 plays an essential role in protecting controversial speech and marginalized groups and viewpoints, as I discussed in response to Senator Cantwell's question above. As historical examples, I would offer several instances involving documentation of violence and state abuse of power. First, filmed evidence of police abuses of Black Americans in the past decade has profoundly changed broader public understanding of this problem. Narrative accounts of such abuse, of course, go back centuries. But Section-230-immunized platforms allowed for filmed evidence of violence against Philando Castile, George Floyd, and other individuals to reach a worldwide audience, and to inspire demands for change. A second example involves platforms' removal of videos documenting police under non-230 legal regimes. To simplify compliance with U.S. copyright law, major platforms may automatically block some uploads containing music at the request of copyright holders--even if the uploads do not infringe copyright. This has led to multiple reported examples in which police played popular songs while being filmed, in order to ensure that footage could not be posted on social media.\23\ --------------------------------------------------------------------------- \23\ BBC, US officer plays Taylor Swift song to try to block video (July 2021), https://www .bbc.com/news/technology-57698858; Dexter Thomas, Vice, New Video Shows Beverly Hills Cops Playing Beatles to Trigger Instagram Copyright Filter (February 2021), https://www.vice.com/ en/article/new-video-shows-beverly-hills-cops-playing-beatles-to- trigger-instagram-copyright-filter/; Dexter Thomas, Vice, Is This Beverly Hills Cop Playing Sublime's `Santeria' to Avoid Being Live- Streamed? (February 2021), https://www.vice.com/en/article/is-this- beverly-hills-cop-playing-sublimes-santeria-to-avoid-being- livestreamed/. --------------------------------------------------------------------------- The final example involves documentation of human rights abuses from the 2010s in Syria. Internet platforms including YouTube, responding to pressure from governments around the world, adopted crude methods to identify and automatically remove potentially terrorist content. Those filters presumably caught and removed a great deal of legitimately harmful and illegal material, including ISIS recruitment videos or images. But they are also widely understood to be the reason for major erroneous removals, including hundreds of thousands of videos collected by the non-profit Syrian Archive as evidence of human rights abuse and war crimes.\24\ --------------------------------------------------------------------------- \24\ Syrian Archive, Removals of Syrian human rights content: May 2020 (May 2020), https://syrianarchive.org/en/lost-found/removals-of- syrian-human-rights-content-may-2020/. --------------------------------------------------------------------------- In these examples, history teaches us that platforms will resort to content moderation efforts that are at best imperfect, and at worst highly dangerous to the voices of victims and marginalized groups. Lawmakers should be cognizant of this risk in considering legal changes. Interoperability: How can interoperability and a middleware ecosystem promote a safer online environment for children or protect users from online censorship? Answer. Interoperability or middleware-based tools for Internet platforms can introduce user control and competition into the online speech ecosystem. I described the basic concept--and some practical constraints on implementation--in a short article.\25\ In the time since that article was published, technical advances have meaningfully lowered some of the barriers to making middleware a reality. Importantly, AI can dramatically lower the costs of content moderation for new market entrants, and can support better models to protect user privacy. --------------------------------------------------------------------------- \25\ Daphne Keller, The Future of Platform Power: Making Middleware Work, 32 J. of Democracy 168 (2021), https:// www.journalofdemocracy.org/articles/the-future-of-platform-power- making-middleware-work/. --------------------------------------------------------------------------- Middleware also offers a more constitutionally viable way forward than many other legislative proposals. A legislative approach that started by lowering current legal barriers to middleware under laws like the Computer Fraud and Abuse Act or DMCA Section 1201 would, to my mind, create no First Amendment issues at all.\26\ A more drastic intervention that actually mandated interoperability would be more complex, but is certainly more constitutionally viable than proposals for top-down state intervention in online speech. As co-counsel and I discuss in a brief to the Supreme Court, interoperability rules would pose far fewer risks to Internet users' First Amendment rights, compared to laws imposing top-down state speech preferences.\27\ Harms to platforms' First Amendment rights would also be reduced, since such laws would not prevent them from setting their own editorial policies for content moderation and ranking. Platforms would likely still argue--correctly--that compulsory interoperation and support for alternate feed offerings are a form of compelled speech, and burden their constitutional rights. But carefully crafted, content-neutral laws might, like the one in the Supreme Court's Turner cases about cable television, still survive a First Amendment challenge in court. --------------------------------------------------------------------------- \26\ See generally Corynne McSherry, EFF, Want More Competition in Tech? Get Rid of Outdated Computer, Copyright, and Contract Rules (2018), https://www.eff.org/deeplinks/2018/12 /want-more-competition-tech-get-rid-outdated-computer-copyright-and- contract-rules. \27\ Amicus brief of Francis Fukuyama, Moody v. NetChoice (2023), https://www.supreme court.gov/DocketPDF/22/22-277/292730/20231207155250803_22-277%20- 555%20Fukuyama% 20Amicus%20Brief%20Final.pdf. --------------------------------------------------------------------------- As Senator Schatz raised in the hearing, many parents in particular do not want to deal with complicated controls. They just want a simple way to protect their kids. I think middleware stands a far better chance of actually achieving that goal than many other approaches. As a parent, I share lawmakers' dissatisfaction with the controls currently offered by major tech companies. The best way to fix that is to allow competitors to do better. At its simplest, this could mean offering parents a one-click option to automate selection among platforms' hard- to-navigate existing controls. That is very much like the middleware product recently created by Professor Ethan Zuckerman to automate user selections among Facebook controls. Zuckerman and the Knight First Amendment Institute argued in court that this tool should be immunized from legal opposition by Meta under Section 230s toolmaker protections.\28\ Far more ambitious tools are also possible, and could offer parents a range of options as well as simplicity of use. --------------------------------------------------------------------------- \28\ Paul M. Barrett, Lawfare, A Zuck Takes on Meta (2024), https:/ /www.lawfaremedia.org/ article/a-zuck-takes-on-meta. --------------------------------------------------------------------------- I do want to be clear on one thing, however. Middleware and interoperability are a way to put users in control of their information diet, instead of governments or platforms. They would not and legally could not prevent those users from choosing to consume lawful misinformation, hate speech, and other harmful material. The hope is that most users would make considered choices, and opt for healthier feeds than the ones they have today. But that is not guaranteed. What is more certain, though, is that viable alternatives would reduce current incumbents' power over discourse. In so doing, it would reduce their resulting grave vulnerability to state cooptation and coercion. Consumer Protections and Privacy Considerations: How would addressing gaps in kids' online safety promote broader consumer protections for all users, regardless of age? Answer. As discussed above in response to Senator Cantwell, I would see grave concerns in legal mandates that require privacy-invasive age assurance or verification. That approach would reduce protections for all Internet users, regardless of age. By contrast, legal reforms that allow middleware and interoperability to address gaps in kids' online safety would benefit adults as well. I do not claim sufficient expertise in U.S. privacy and consumer protection law to weigh in meaningfully on most other current gaps or their potential corrections. To the extent that consumer protection means basic non-content-related privacy protections or the avoidance of unfair and deceptive practices, however, I do not see Section 230 or the First Amendment as barriers. Problems would arise, however, if such laws became a de facto mechanism for state control over online speech. That could happen, for example, if Federal agencies were called upon to interpret inevitable disputes or gray areas in platform content rules. It could also happen under well-intentioned transparency laws that inadvertently unleashed similar state power over speech. I discuss a few brief examples of that problem and analyze First Amendment issues in footnoted materials.\29\ --------------------------------------------------------------------------- \29\ Daphne Keller, CIS Blog, State Abuse of Transparency Laws and How to Stop It (2022) (examples), https://cyberlaw.stanford.edu/blog/ 2022/09/state-abuse-transparency-laws-and-how-stop-it/; Daphne Keller, Platform Transparency and the First Amendment, 4 J. Free Speech L. 1 (2023) (legal analysis). --------------------------------------------------------------------------- ______ Response to Written Questions Submitted by Hon. Edward Markey to Daphne Keller Section 230 and Jawboning: Republicans have argued that the Biden administration coerced social media platforms to take down misinformation, hate speech, and other content with which it disagreed. I strongly disagree with those claims. In making these arguments about jawboning, Republicans conveniently ignore the yearslong campaign among congressional lawmakers to pressure social media platforms into changing their content moderation policies to be ``fairer'' or ``less biased'' against conservatives. Can a member of Congress violate the First Amendment by using their legislative authority to coerce a third party to take action against speech of a private citizen? Answer. This is a complicated question. Members of Congress could certainly cross the line into unconstitutional coercion, including through pressuring platforms. They could also violate the First Amendment by a means raised in circuit court precedent but not discussed in the recent Supreme Court jawboning cases: by giving instructions to a fully compliant platform, which willingly silences user speech at state behest. Such state cooptation of platform power is just as big a threat as state coercion. At the same time, as I have discussed elsewhere, not every government communication--and not even every heated or angry statement--is jawboning.\30\ Open and frank discussion between platforms and those in government is essential for sound policymaking. It would defeat key functions of democracy if our representatives, once in office, could no longer express the beliefs and policy preferences that got them elected in the first place. I believe that the constitution gives Members relatively wide leeway to criticize platforms and argue for changes. Notably, individual Members also have limited ability to credibly threaten legislative change, since no Member can--acting alone--follow through on that threat. --------------------------------------------------------------------------- \30\ Daphne Keller, Knight First Amendment Institute Blog, Six Things About Jawboning (2023), https://knightcolumbia.org/blog/six- things-about-jawboning. Could an open-ended congressional investigation, including through the use of subpoenas and other compulsory tools, into third-party entities because of their speech or others' speech that they distribute or publish constitute unconstitutional jawboning? Answer. Abuse of Congress's investigatory powers, through subpoenas and other means calculated to chill speech or retaliate against speakers, can violate targets' constitutional rights. The precise legal tests are unclear, however, and recent precedent on this legal issue is sparse.\31\ Intermediaries ranging from platforms to law firms to universities to broadcasters can have powerful incentives to comply with government demands and appease those in power. Relatively few take legal measures to resist Congressional inquiries and create new case law.\32\ Victims of this appeasement strategy are likely to be unaware of the government's role in depriving them of legal representation, publication channels, and other means of exercising First Amendment rights. --------------------------------------------------------------------------- \31\ Christopher Dunn, NYCLU Blog, Constitutional Limits on Congressional Investigations Into Palestine-Israel Advocacy (2024), https://www.nyclu.org/commentary/column-constitutional-limits-on- congressional-investigations-into-palestine-israel-advocacy. \32\ See Emma Llanso and Taylor Moore, CDT, Senate Inquiry into Backpage's Content Moderation Practices Would Set Dangerous Precedent for Free Speech Online (2016), https://cdt.org/insights/senate-inquiry- into-backpages-content-moderation-practices-would-set-dangerous- precedent-for-free-speech-online/. --------------------------------------------------------------------------- Attention to Congressional abuse of investigatory powers was acute during and immediately after the mid-twentieth-century McCarthy era. Professor Seth Kreimer, in his seminal 2006 Censorship By Proxy article, pointed to these Congressional excesses as major drivers of the same mid-century Supreme Court rulings that inform our understanding of jawboning and platform censorship today.\33\ These include New York Times v. Sullivan (about newspaper liability for a third party advertisement by civil rights activists), Bantam Books (about state pressure on booksellers), and Smith v. California (about bookseller liability under law). The throughline Professor Kreimer sees is a judicial rejection of claims that ``indirect'' state pressures were ``immune from constitutional scrutiny because their impact was mediated by private implementation.'' --------------------------------------------------------------------------- \33\ Seth Kreimer, Censorship by Proxy: The First Amendment, Internet Intermediaries, and the Problem of the Weakest Link, Penn Law School Public Law Working Paper No. 06-45 (2006), https://ssrn.com/ abstract=948226. --------------------------------------------------------------------------- In cases about that era's legislative abuses, the Supreme Court noted that ``[t]here is no doubt that legislative investigations, whether on a Federal or state level, are capable of encroaching upon the constitutional liberties of individuals.'' \34\ As a result, the ``power of compulsory process [must] be carefully circumscribed when the investigative process tends to impinge upon such highly sensitive areas as freedom of speech or press, freedom of political association, and freedom of communication of ideas, particularly in the academic community.'' In another case, it observed that the threat of compelled legislative testimony might chill potential witnesses from expressing any but the ``most orthodox and uncontroversial views[.]'' \35\ The fact that such pressure came not directly from legislators but from the ``glare of publicity'' and resulting ``public stigma, scorn and obloquy'' did not, the court held, relieve state actors of their own constitutional responsibilities. Today's victims of pressure campaigns may not know the word ``obloquy,'' much as the justices who authored these rulings would likely not have recognized words like ``doxing.'' The fundamental constitutional concerns, however, remain unchanged. --------------------------------------------------------------------------- \34\ Sweezy v. New Hampshire, 354 U.S. 234 (1957). \35\ Watkins v. U.S., 354 U.S. 178 (1957). To be more specific, congressional Republicans have repeatedly suggested that Congress should pass legislation to reform Section 230 so that only platforms that meet a standard of ``neutrality'' will receive Section 230 immunity. Could such legislation or the threat of passage of such legislation, especially by members or leaders of congressional committees with jurisdiction over Section 230 and the Telecommunications Act, encourage or pressure platforms to modify their content moderation practices? If so, would this conduct constitute unconstitutional jawboning? Answer. I do not believe that current law provides a clear answer to this question. My best take--which other experts might dispute--is that the behavior described would typically not constitute unconstitutional jawboning. Predicating Section 230 immunity on ``neutrality'' would, to be clear, be a catastrophically bad policy choice. It would likely face a First Amendment challenge as an ``unconstitutional condition.'' But threatening to pass flawed or even unconstitutional laws is a Congressional prerogative. Ideally, bad legislative proposals are one step in a deliberative and iterative process leading to good ones. The scenario described raises some red flags, but nonetheless does not involve some of the key indicia of unconstitutional jawboning. Members are not offering a bargain in which they will stay their hands if platforms suppress particular speech. By contrast, the behavior held unconstitutional in the Supreme Court's recent Vullo ruling involved an explicit quid pro quo. The Members do not suggest that platforms are violating existing laws, or threaten to carry out or instigate enforcement, as was the case in both Vullo and Bantam Books. Any implied threat of retaliation is weakened by the fact that individual Members cannot unilaterally enact new laws, as the Ninth Circuit noted in a jawboning case brought by Robert F. Kennedy Jr. against Senator Elizabeth Warren.\36\ --------------------------------------------------------------------------- \36\ Kennedy v. Warren, 66 F.4th 1199 (9th Cir. 2023). The threats are made publicly, which somewhat reduces the suggestion of a backroom deal or coercion, and enables public review and accountability--whether in court or at the ballot --------------------------------------------------------------------------- box. I wish I could offer a single crisp legal test for jawboning, but I do not believe one exists or is likely to. The potential variables in each situation are too great.\37\ But I think this is the answer that hews most closely to existing case law, and that preserves adequate room for trial, error, and debate in the legislative process. --------------------------------------------------------------------------- \37\ I list those factors, as I see them, in the final section of a blog post, Daphne Keller, Knight First Amendment Institute Blog, Six Things About Jawboning (2023), https://knightcolum bia.org/blog/six-things-about-jawboning. --------------------------------------------------------------------------- ______ Response to Written Questions Submitted by Hon. Maria Cantwell to Nadine Farid Johnson Carr Pressure and Jawboning: Earlier this month, FCC Chairman Brendan Carr once again threatened to revoke broadcast licenses for press coverage that the President does not like. This time it was because the press is reporting the facts about the Iran war. What's unsettling is that we have seen this type of behavior from Chairman Carr before--whether it's political commentary on late-night or reporting about ICE raids in a community. I agree with Chair Cruz that this type of coercion is ``dangerous as hell.'' It is a threat to free speech and the freedom of the press. Question 1. Do you agree that a sitting government official using the threat of license revocation to suppress news reporting he and the President don't like represents one of the most serious threats to the First Amendment and free speech that we face today? Answer. Chairman Carr's threats of regulatory reprisal in retaliation for constitutionally protected expressive decisions, including editorial decisions about how to report on factual information concerning a matter of national security, appear to be government censorship--an unconstitutional effort by the government to use coercion to suppress speech it disfavors. The government may not seek to suppress speech it disfavors, including by jawboning--an act of government pressure to suppress constitutionally protected speech. As the Supreme Court held in Bantam Books, Inc. v. Sullivan, the First Amendment bars the government from coercing private speech intermediaries into suppressing speech it disfavors. The point at which government efforts to pressure a speech intermediary cross the line into unconstitutional coercion is a vitally important question. This is because some efforts by government actors to engage with speech intermediaries are legitimate as an aspect of governance, including a manifestation of the public's right to hear its government's position. Efforts to coerce, however, are unconstitutional, because they leverage the power of the state to censor public discourse. Put simply, the government may not indirectly restrict speech that it may not restrict directly. This problem of government coercing private entities to revise their speech-related practices is extensive and continues to broaden, encompassing not only media outlets but also technology platforms, universities, and more. Irrespective of the entity targeted, and irrespective of the Administration responsible, government coercion to suppress speech is unconstitutional and presents an existential threat to the ability of the public to say what they wish and to hear what their fellow citizens say. Congress should act to make sure that governmental efforts at coercion do not evade public and judicial scrutiny. Section 230 Liability Protection: Section 230 was adopted thirty years ago on a simple idea: platforms hosting user speech should not be treated as publishers of every post, and those that moderate harmful material should not be punished for doing so. That framework has been critical for startups, small businesses, and online communities that depend on the ability to host user content without being sued over every post. But thirty years of experience have also shown that the largest platforms used these protections very differently from how Congress intended. They built business models around algorithmic amplification, addictive design, and pervasive data collection. They hid behind Section 230 while their products harmed children and disrupted elections. As we consider Section 230 at thirty, I want to understand what has worked well and what hasn't. Specifically, what these protections have meant for free speech and for the smaller platforms that depend on them the most. But also how we hold tech platforms accountable for their harms. Question 1. Do you agree that large platforms need to be held accountable for their actions and the harms they are causing to society--particularly harms to children and data privacy? Answer. Outside of a few narrow exceptions, the First Amendment protects even speech that many would deem outrageous, harmful, and dangerous. That is why, even without Section 230, the platforms likely could not be held liable for the vast majority of speech that they host that has been the focus of public concern. The better path to platform accountability lies in legislation that is both aligned with First Amendment rights and would foster a better online environment for people who use social media platforms, including children. Congress should enact new mandates in transparency, privacy, and interoperability. Legislation focused on such structural changes-- including comprehensive data privacy legislation--would serve to minimize the harms of social media and liberate users from the monopoly control over their online speech. ______ Response to Written Question Submitted by Hon. Amy Klobuchar to Nadine Farid Johnson Platform Accountability and Transparency Act: I support the Platform Accountability and Transparency Act, with Senator Coons, which would require social media companies to share data with researchers and the public. Question 1. Why is it so important to give researchers access to information and data from platforms to understand how they shape society? Answer. The Knight Institute greatly appreciates your support of the bipartisan Platform Accountability and Transparency Act led by Senator Coons. Since 2018, the Knight Institute has been advocating for a safe harbor for public-interest research and journalism focused on social media platforms. The Institute published a white paper proposing a safe harbor for platform research in 2022, and supported the introduction of a previous version of the Platform Accountability and Transparency Act that same year. Independent research into and public understanding of social media platforms is vitally important because of the ways these communications platforms have transformed our digital public sphere and the outsized role they play in our democracy. Without this transparency, the public--and policymakers--are left with only the information about the platforms that is provided by the platforms themselves. But journalists and researchers who study the platforms and their black-box algorithms continue to do so under the threat of serious legal liability. This is because the companies' terms of service generally ban the use of the tools necessary to study the platforms at scale, subjecting researchers to the threat of civil and potentially even criminal liability for public-interest investigations. A legal safe harbor for those who study the platforms in ethical and privacy-preserving ways is critical to improving public understanding of how the platforms are shaping society. ______ Response to Written Questions Submitted by Hon. Brian Schatz to Nadine Farid Johnson Section 230s Impact on Censorship: Question 1. Some experts have raised concerns that repealing Section 230 could result in platforms over-moderating content, which may disproportionately impact content deemed ``controversial.'' What content would be most vulnerable to over-moderation and censorship? How might these online harms translate to the real world? What has history shown us? Answer. Because the First Amendment protects harmful, outrageous, and sensational speech, the platforms would not be required to remove it, even without Section 230. But they would be motivated to take down speech that might plausibly give rise to liability. That category would include, most significantly, potentially defamatory speech, such as speech that alleges wrongdoing by a specific individual. The First Amendment does not protect defamation. Truth is of course a defense to defamation liability, but the platforms do not have the ability to determine--definitively and at the scale at which they operate--whether any particular allegation is true or not. This means that, absent Section 230, the platforms could face liability for leaving up factual allegations about real people, and so they would almost certainly remove them, including allegations that turn out to be true. Needless to say, true allegations of wrongdoing--particularly allegations about government officials and powerful private actors--are socially valuable and important for public discourse and a healthy, accountable democracy. Incentivizing platforms, many of which use LLMs or other automated tools for content moderation, to remove anything potentially defamatory, would in that sense result in a less informed citizenry and less accountable powerful actors. There is a clear analogy to the implementation of the Digital Millennium Copyright Act (DMCA), which requires that platforms remove content that infringes on copyright. In practice, without the ability to review every takedown request from self-described copyright holders for fair use or other exemptions, platforms have erred on the side of removing any material for which they receive a takedown notice--often through automated processes--which has frequently resulted in the removal of non-infringing content.. There is good reason to think that platforms would approach potential legal liability for defamation the same way, playing it safe by casting a very wide net on what they take down, almost certainly removing not just legal but also socially valuable speech in the process. Powerful actors are likely to retain well-resourced legal counsel specifically to pressure platforms to remove allegedly defamatory content about them--making this problem all the more acute. Transparency: Question 1. Please elaborate on how an approach that conditions Section 230 on strong accountability measures can reform platform behavior without suppressing speech. Answer. Because Section 230 affords protections above and beyond what is required by the First Amendment, the Knight Institute believes it would be consistent with First Amendment principles to condition those protections on large platforms' compliance with certain mandates, so long as the conditions are themselves constitutional. There are important First Amendment limitations on the government's ability to restrict disfavored speech by either the platforms or their users, and the Institute would oppose any attempt by Congress to legislate around those limitations at the expense of constitutionally protected speech. We believe it is possible, however, to enact structural reform that is consistent with the First Amendment and that would serve to better align the incentives of the large platforms with the interests of their users. Specifically, Congress should enact new mandates in transparency, privacy, data portability, and interoperability. Doing so holds the promise of simultaneously minimizing the harms of social media and liberating users from the monopoly control over--and suppression of-- their online speech. This can be done in a way that not only respects the limitations of the First Amendment, but actually promotes the values that underlie it. While we believe these mandates could and should stand alone (so that they extend to actors not relying upon Section 230), Congress could also condition Section 230 protections--at least for the largest companies with the resources necessary--on compliance with mandates such as researcher access, data privacy, and user controls. This is particularly so for the largest social media companies, whose growth has been made possible in large part by Section 230s expansive protections. Conditioning those protections on transparency, privacy, and interoperability mandates would protect the speech rights of platforms and of their users, while also incentivizing the platforms to adapt to a more competitive landscape. Question 2. How can improved transparency measures, such as what is included in the Internet PACT Act, enhance trust between users and platforms? Answer. Independent research into and public understanding of social media platforms is vitally important because of the ways these communications platforms have transformed our digital public sphere and the outsized role they play in our democracy. Without this transparency, the public--and policymakers--are left with only the information about the platforms that is provided by the platforms themselves. Giving users the information they need to make informed decisions about which platforms to use, what personal data to share, and what their options are should they encounter harmful or illegal content, could help to promote a more functional digital public square. Question 3. How can safe harbors for researchers promote positive outcomes for online safety and transparency? Answer. Journalists and researchers who study the platforms and their black-box algorithms continue to do so under the threat of serious legal liability. This is because the companies' terms of service generally ban the use of the tools necessary to study the platforms at scale, subjecting researchers to the threat of civil and potentially even criminal liability for public-interest investigations. A legal safe harbor for those who study platforms in ethical and privacy-preserving ways is critical to improving public understanding of how the platforms are shaping society. ______ Response to Written Questions Submitted by Hon. Amy Klobuchar to Matthew Bergman Section 230 Sunset: Question 1. How would repealing Section 230 create real incentives for platforms to design their platforms in a way that they are safe for users? Answer. Repealing Section 230 is neither necessary nor appropriate. Rather, Congress should simply reaffirm its original intent set forth in the plan language of Section 230 that It is the policy of the United States-- (1) to promote the continued development of the Internet and other interactive computer services and other interactive media; (2) to preserve the vibrant and competitive free market that presently exists for the Internet and other interactive computer services, unfettered by Federal or State regulation; (3) to encourage the development of technologies which maximize user control over what information is received by individuals, families, and schools who use the Internet and other interactive computer services; (4) to remove disincentives for the development and utilization of blocking and filtering technologies that empower parents to restrict their children's access to objectionable or inappropriate online material; and (5) to ensure vigorous enforcement of Federal criminal laws to deter and punish trafficking in obscenity, stalking, and harassment by means of computer. 47 U.S.C. Sec. 230(b). The expansive application of Section 230 is not the product of statutory intent but of judicial interpretations that ignore Congress' express policy. With this broad construction of Section 230, Internet providers have been protected from liability even though they republished content knowing it might violate the law, encouraged users to post illegal content, and changed their design and policies for the purpose of enabling illegal activity. And, predictably, tech companies have seized upon this trend to evade all possible legal accountability simply because third-party content is found somewhere in the causal chain of their misconduct. The expansive interpretation of Section 230 far beyond Congressional intent has allowed tech companies to escape accountability for decisions that harm the users of the very platforms they design, market, and operate. Change will only come when they must bear the economic costs of their deliberate design decisions. Restoring Section 230 to its original intent would accomplish just that. These companies would be forced to internalize the cost of safety when designing their platforms and may finally opt for safety instead of growth. If the designers of the dangerous platforms face the consequences instead of the families of those who use them, then they will have the incentive to change their behavior. Through the imposition of civil liability, we can change their economic calculus. For example, safer designs, effective age verification to prevent children from accessing their platforms, and parental controls that allow parents to effectively protect their children may outweigh maximizing time spent at all costs. Conforming Section 230 to its original purpose would not to impose any special duty on tech companies. Instead, it would even the playing field and require these companies to abide by the same rules that ever other company is subjected to and follow the same incentive structure that every other company must assess when designing their products. Every other company in America operates under a duty of reasonable care. Social media and AI companies should be held to that same standard. ______ Response to Written Questions Submitted by Hon. Brian Schatz to Matthew Bergman Design Choices and Product Liability: Question 1. Please elaborate on how harmful design choices fall outside of the scope of Section 230 protections. Answer. Although Section 230 presents a substantial barrier to seeking justice for harms done by technology platforms, some courts have recognized the protections of the law should not extend to some design decisions that are completely separate from third-party content. For example, filters created by the social media platforms do not involve third-party content and are thus outside the scope of the vast protections of Section 230. The same is true for notifications sent to users, often throughout the night and during school hours, in an effort to pull them back into the platforms. Features such as likes or subscriber count hook the user, especially minors, often creating a lack of self-worth if the user doesn't receive enough. Ephemerality creates a sense of ``fear of missing out'' for young users, prompting them to get back on the platforms more often so as to not miss anything. Many of these features were designed in furtherance of the companies' overall goal to addict users, including children, to their platforms. Indeed, increasing the time spent on the platform was at the forefront of the social media executives' minds, without regard to the cost in may extract from young users. In fact, addicting children wasn't just a byproduct, it was the goal. As a Meta internal document shows, ``tweens'' were sought out as users, in an effort to hook them while they are young and ensure they use the products into their ``teen'' and adult years. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In 2015, Meta CEO Mark Zuckerberg instructed company executives that he hopes to accomplish an increase on 12 percent in the amount of time users spent on the platforms. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] One internal e-mail states that Mark Zuckerberg ``has decided that the top priority for the company in 2017 is teens[.]'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In a 2016 internal presentation, YouTube declared that the ``vision'' for the YouTube Kids platform was to ``create the next generations of Youtube Users'' and to make using YouTube a ``part of the daily routine for all kids under 13.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Despite vehemently arguing that Section 230 is an all-encompassing safeguard (it is not), big tech companies may be held liable for creating dangerous social media products, purposely addicting children to these dangerous platforms, and failing to warn about the harms that may result therefrom. Question 2. Do social media companies have a responsibility to prioritize the wellbeing of children through their design choices? What are the consequences of inaction? Answer. When it was introduced, the primary purpose of the Section 230 was to protect children from sexually explicit Internet content. As the late Chief Judge Katzmann of the Second Circuit recognized, ``The text and legislative history of [Section 230] shouts to the rafters Congress's focus on reducing children's access to adult material.'' Congress' express purpose in enacting Section 230 was ``to encourage the development of technologies which maximize user control over what information is received by individuals, families, and schools . . . empower parents to restrict their children's access to objectionable or inappropriate online material; and ensure vigorous enforcement of Federal criminal laws to deter and punish trafficking in obscenity, stalking, and harassment by means of computer.'' 47 U.S.C. Sec. 230 (b)(2)-(4). Congress further sought to provide ``[p]rotection for `Good Samaritan' blocking and screening of offensive material,'' by providing that ``no provider or user of an interactive computer service shall be held liable on account of any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether such material is constitutionally protected.'' 47 U.S.C. Sec. 230 (c)(2)(A). However, despite the purpose of Section 230, companies have continued to ignore the wellbeing of children when designing their products, hoping to escape any potential liability by using Section 230 as a shield. The overly expansive interpretation of Section 230 among some courts has led to the dismissal of meritorious cases, further supporting these companies' belief that they may continue to design and operate their platforms in a dangerous manner, leading to the harm and, in some cases, death of children. And they may continue to do this without fear of liability without major reform to the law they have hidden behind for thirty years. We must remember that Section 230 was created in the early days of the internet. before social media and before AI. The consequences of the text could not be seen at that time, as it was the age of chat forums and message boards, but they can be seen now. The consequences are borne by the children who use these platforms and their families, who are burdened with the realities of attempting to help a child who has been harmed by some of the most powerful companies in the world. Without change, families will continue to be torn apart while powerful executives reap the profit gained by harming children. Question 3. Late last year, the public learned in unsealed court briefings that social media companies may have delayed design changes that could have protected children sooner. How would the reinterpretation of Section 230 by courts change incentives for social companies to put children first? Answer. Consider the recent verdict coming out of a Los Angeles courtroom. My firm's client, known by K.G.M., was awarded $6 million dollars by a jury based on harm done to her by Meta and YouTube, in association with her use of Instagram and YouTube. The jury found that both companies had been negligent in their design and operation of each respective platform, and that their conduct warranted an award of punitive damages. However, throughout years of litigation in that case, the social media companies insisted that Section 230 protected their conduct. And although the court there rejected the companies' arguments that every single action was protected by Section 230, it did find that some particular conduct by the defendants or features on the social media platforms could not serve as a basis for liability because of Section 230. K.G.M.'s verdict is the first of its kind and may serve to help incentivize these social media companies to put children first on occasion, as they have to consider that some cases may pass the Section 230 hurdle. However, a reinterpretation of Section 230, would further balance the scales and incentivize companies to change their behaviors and practices. Social media companies, without the version of Section 230 that exists today, would no longer be allowed to needlessly protract litigation and delay plaintiffs' getting their day in court. ______ Response to Written Questions Submitted by Hon. Ben Ray Lujan to Matthew Bergman Question 1. In your litigation experience, how are social media companies weaponizing design decisions to keep kids addicted to social media? Answer. Social media platforms embody numerous design features that promote addiction to their platforms. This includes design features that exploit intermittent reward, social comparison metrics, and ``flow'' state. Each of these may discussed in turn, but they all work in tandem to keep users hooked and have their foundations in behavioral psychology. Think of slot machines, where gamblers cannot predict which pull will result in payouts, and as they play, they are constantly hearing others around them win, reminding them that their next pull might be a jackpot. All three of these addictive design features are built in to social media sites and are reflected in the ``hook model'' studied by Meta's researchers: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In a 2015 internal presentation, Google employee Tristan Harris noted that ``[i]ntermittent versus predictable rewards are the most addictive and hardest to stop. It's why slot machines make the most money at casinos'' and that ``[t]hese are attention casinos[.]'' He proposed instead that Google design its platforms to ``minimize the presence of intermittent variable rewards and, and reduce addictions'' but his concerns were ultimately disregarded by the company. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This figure, which is taken from a book by Nir Eyal titled Hooked: How to Build Habit Forming Products, also appears in Snap's documents. The figure is annotated as follows [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] A. Likes, Comments, and Other Metrics Scientists have found that receiving ``likes'' on social media platforms are very similar to the ``rewards'' that researchers have associated with addiction research for decades. Likes can include the ``like'' button on Facebook and YouTube but can also include ``hearts'' on Instagram and TikTok. Similar metrics include the number of shares on a post, number of comments, and number of followers or friends for a user. While seemingly innocuous, users' quest for these publicly visible ``rewards'' has been linked to a number of secondary harms such as reduced sleep efficiency and duration due to ``routine check[ing]'' behaviors during the night. They have also been associated with reinforcing addictive behaviors in order to encourage users to spend more time on these platforms. They have even been linked to increased feelings of depression, anxiety, and negative social comparison since receiving ``fewer likes'' is viewed as a form of ``negative peer feedback.'' Comments are similar to likes in that they provide users with ``quantifiable (and qualitative) feedback'' about their experiences on the platform. For teens, this means that they are able to quantify the ``success'' of their posts with many teens reporting that they post ``self oriented images on social media with the goal of obtaining likes and other forms of feedback such as comments.'' Other studies have found that comments are often seen as a mechanism for ``gaining [social] status'' and reflect a degree of ``digital social approval.'' As a result, other studies have linked comments to negative social comparison since ``not receiving enough likes on one's pictures can negatively affect appearance esteem and prompt delet[ion] of a post.'' B. Algorithmic Recommendations At the heart of every social media platform's engagement strategy is their proprietary recommendation algorithms. These algorithms are optimized to maximize time on the platform rather than healthy interactions with a person's social network. They accomplish this in a variety of ways. One way they drive time spent on social media is by inferring the interests of the user (which may or may not be expressed by the user in any direct way) and feeding them an aggregation of posts that, while most likely to keep them engaged, may lead the user down problematic rabbit holes. Regarding TikTok, these problematic rabbit holes were also known as ``negative filter bubbles.'' A 2021 TikTok presentation entitled ``Negative Filter Bubble Project Sharing'' noted that ``Algorithms lead[] people into a feed filled with continuous similar negative videos that are hard to escape.'' What's more, another internal document notes that ``minor creators [fall] into higher density of negative comment filter bubbles.'' To its credit, TikTok did set as a 2023 safety goal to ``be more transparent to the public and trusted by teen users and parents,'' in addition to building ``industrial leading teens safety and well-being features.'' As discussed below, these aspirations of transparency and improved safety features that were noted approximately six years after TikTok launched in the United States were too little and too late for many children and teens. Another way recommendation algorithms drive time spent on social media is by utilizing the intermittent variable reward mechanism that Skinner discovered (e.g., likes, notifications, comments that appear unpredictably), which contribute to addiction to the platforms themselves. The addictive power of an engagement-optimized algorithm is recognized within the medical and academic literature. In the Handbook of Children and Screens, under the chapter regarding ``Problematic Internet Use,'' the authors' consensus was that ``Attention focused designs intended to generate, or possibly exploit, potentially addictive features (e.g., `likes') and conditioned responses (e.g., notifications) alongside powerful algorithm-based technologies may lead youth to stay online longer than either intended or recommended.'' C. Auto-Play Academics recognize that autoplay is a feature that is designed to prolong engagement at the cost of displacing important developmental opportunities for young children and is even associated with greater child behavioral difficulties. It has also been linked to difficulties controlling device use more broadly. The latter phenomenon has even been observed in social media companies' own research. For example, Meta's researchers learned that clinicians regarded autoplay as not having any ``beneficial'' role while ``detract[ing] from patients' ability to control amount of time spent'' using their platforms.'' A YouTube study found that autoplay contributed to increased nighttime watchtime between the hours of 12:00AM and 6:00AM and that ``autoplay does affect watch behavior at night and is potentially pushing users into bad sleep habits.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] It specifically noted that ``disabling autoplay late into the night could potentially help in adding 1M hours in sleep time in California everyday!'' However, no such disabling ever occurred. D. Infinite Scroll Like autoplay, the Infinite or Endless Scroll feature has also been linked to prolonging engagement at the cost of displacing children's developmental opportunities. This feature is also linked to behaviors that create a user-sided ``time distortion'' that results in users spending more time on social media platforms than they originally intended. Some studies cited (and commissioned) by the social media companies have found that these effects are mitigated by ``active'' use of their platforms--e.g., generating content or posting content. However, this hypothesis has not been widely accepted, particularly since there is evidence that children are more likely to ``watch, play, or scroll through content created by others than they are to use their devices to produce their own content.'' E. Notifications Meta studied notifications and the ability for this feature to induce habitual or addictive behaviors. For example, one internal document supports the basic addiction principle that ``experiencing a reward (or reinforcement) can increase learning and motivation. This contributes to repeated, potentially habitual behaviors.'' The document included the following figure: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] It is not surprising then, that among the many metrics Meta tracks, the ``success rate'' of their notifications, defined as increasing daily usage, is a key one. Darius Kilstein, a Director in Data Science at Instagram, reports: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] This is to say that Meta is monitoring (and presumably modulating) the intermittent reward mechanism to ensure that engagement is maximized. An internal YouTube presentation noted that YouTube's ``stickiness'' was, in part, attributable to receiving a notification about likes and comments on a video. It noted that a likes and comments notification ``causes users to feel that they must be aware of what is happening on the platform'' and it ``keeps users on the platform longer[.]'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In fact, YouTube noted that ``notifications are a critical part of YouTube and contribute to addiction.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] F. Beauty Filters Despite being a relatively new social media feature, Augmented Reality (``AR'') filters--commonly referred to as ``Beauty Filters''-- have been thoroughly studied due to the outsized negative impact they could have on users by exacerbating a socio-psychological phenomenon known as ``social comparison.'' Prior to the advent of these Beauty Filters, researchers had already identified that social media may exacerbate social comparison and lead to increased body dissatisfaction and/or disordered eating. However, with the introduction of Beauty Filters--many developed by social media companies--users are now exposed to ``manipulated'' photos that depict unrealistic (and in some cases impossible) body image Standards. In an internal e-mail chain, Meta executives discussed a decision to reverse the ban on Instagram beauty filters that resembled cosmetic surgery. In one e-mail, an executive urged others to rethink this decision, stating that it would be putting growth over responsibility. [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] G. Internal Company Documents Show They Knew What They Were Doing Notably, many of the features recognized in the literature as addictive and/or harmful were likewise identified in research conducted by Meta and presented in 2021. A Mixed Methods Clinicians study identified product features and pathways impacting mental health, including the following: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Of these features, several were identified as ``primarily negative,'' including Video/Photo filters, location sharing, automatically playing videos, and pop-up notifications. Internally, social media companies' documents recognize that they could increase engagement by changing the design of the social media. As early as 2016, Meta (then Facebook) was exploring ways to keep teens on its site and posting content: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] As Max Eulenstein, Vice President of Products, said in a Meta e- mail on January 26, 2021, ``No one wakes up thinking they want to maximize the number of times they open Instagram that day. But that's exactly what our product teams are trying to do.'' One mechanism that appears to be especially effective at engaging teens (and adults) is the use of ``reels'' or short videos that repeat. Meta adopted reels from TikTok after seeing how effective they were at promoting usage. Below are Instagram's metrics on the viewing of such reels by teens: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The data demonstrated that active daily U.S. teen Instagram users viewed on average almost 105 reels per day on the platform for a total of 22.5 minutes per day or an average of 20 seconds per reel per day. Briefly watching a short snippet algorithmically curated to one's interest is gratifying via the dopamine reward pathway. Further, their internal research identified the following ``Triggers for Problematic Use on Facebook'' that were later included as part of a ``Problematic Use Journey Map'' in 2020: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Each ``opportunity'' for a reward represents a potential algorithm tweak, and many--if not all--of these were eventually incorporated in some form into the platform. For example, a Facebook presentation contained following: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta's more recent documents depict a Venn diagram portraying one way to conceptualize the usage experience from a design perspective: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] ``Dark Pattern'' is a user experience term referring to interactions that are deceptive, or that trick you into doing something you didn't want to do. Again, this diagram acknowledges that Meta uses ``addictive'' design features and that some of them are set to ``default.'' Meta documents from Project ``Plato,'' intended to study and mitigate ``dark patterns,'' state that ``[s]ome UX [user experience] patterns rob users of their agency.'' Included among them are: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta's researchers also knew that being too critical of these design choices could lead to criticism from internal stakeholders. For example, Jennifer Guadagno noted that her wellbeing team's efforts to study Facebook addiction as part of Project Plato could ``get heavy pushback internally'' and added that she was not ``sure if it'll even be allowed to happen'' because ``if we now know all these things that are potentially bad and then we don't do anything to fix them'' it could lead to problems for the company externally. Arturo Bejar testified, in deposition at in open court, that Meta did not do enough to warn parents or curtail problematic use of their products and that the word addictive is `radioactive.' '' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Despite learning of these problems as early as 2018, Meta had not made meaningful changes to the platform nearly two years later. A Meta presentation from 2020 arrived at nearly identical conclusions as the Project Plato researchers, including the connection between design features and problematic usage. One slide from that 2020 presentation stated that research participants reported ``10+ triggers contributing to [problematic use] habits'' including: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Meta was not the only social media company to reach these conclusions about its platform features. For example, YouTube describes ``finding a video on YouTube search'' as a ``predictable reward'' compared to ``unpredictable rewards'' such as ``finding a new favorite song while in a mix,'' which the company described as ``disproportionately [more] delightful compared to predictable rewards because they're unexpected or exceed expectations.'' YouTube's researchers explaining the reason for this dichotomous design have stated: When thinking about habit building around YouTube, it is important that we reliably fulfill their goal pursuits (reliable reward) while also providing surprising nuggets of reward (variable reward). Ultimately, you're giving people more reasons to come back until they can't even remember why they did. When is the last time you had a goal in mind when you went to Facebook?'' Internal documents also reveal that Instagram is tracking usage/ engagement by teen users. An e-mail from Darius Kilstein, on February 5, 2022, states that Meta ``looked into the long-term decline of feed imp[ressions] for Teens,'' discovered that ``teens aren't relying on Feed for interest consumption as much as they used to,'' and hypothesized that this shift might explain ``why teens now consume fewer Reels than adults.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The reported ``headwind'' for the ``cohort'' points to Instagram's business need to redesign features to better engage younger children and keep up with their competitors (particularly TikTok), as Kilstein explained later in the same e-mail. The ``infinite feed'' invented at Facebook was exported and adapted to Instagram. In her deposition, Dr. Alison Lee stated: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The quote begins with the scenario whereby someone has been ``scrolling'' all day (itself an implicit acknowledgement of overuse) and exhausts their ``connected'' feeds at which point a new, end of feed algorithm is triggered. Aza Raskin is widely considered to have invented the infinite scroll. This feature allows a user to scroll indefinitely through their feed, receiving endless posts and intermittent variable rewards. Raskin testified that infinite scroll was like ``digital cocaine'' for the user. One of the technology leads for Instant Articles on Android at Facebook reported ``we only care about things like time spent, open links etc. That's what we optimized for. That's what we defined as success and failure.'' In 2022, Meta launched Project MYST, a prospective, longitudinal study to follow 4000 teens (13-17 years of age) using the highly respected National Opinion Research Center (NORC) at the University of Chicago for survey data and Meta logs of activity on their platform. Using validated measures for a variety of outcomes and actual usage patterns on their platforms, MYST appeared by all accounts to be a robust, well-designed longitudinal study--the kind that only industry could conduct based on their access to actual social media use data. Among other things, MYST intended to test the following negative hypotheses: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] While the initial plan for the study contemplated its results being released publicly, there has been no evidence that Meta ever publicly acknowledged that the study was conducted or released any results prior to it being produced in the course of litigation and revealed to the public as part of the K.G.M. civil trial in Los Angeles just this past month. Indeed, as the study moved forward, Meta employees seemed to become explicitly fearful of it being publicly released. Leaders directed researchers to refer to the study only by its full name, because they would ``really prefer not to have a cute acronym (MYST). That makes things sound more interesting when/if it leaks. Can we please keep our names as boring and uninteresting as possible.'' Perhaps more illustrative is the tension that is recognized in Meta's internal documents discussing MYST, which reflect that ``there is a fundamental tension between doing rigorous research to identify whether there are potential opportunities for change/intervention and advancing external/narrative goals.'' This tension with external goals is reflected in Meta's many decisions to not share publicly what their research or outside experts conveyed internally. External goals included the goal to ``[a]dvance the credibility of research that finds small or null correlations between social media well-being and/or well- being relative to other factors,'' and to ``deflate the conversations about research claiming causal connections between social media and mental health and well-being.'' Perhaps most critically, Meta's stated external goal was to ``[m]ake the issue bigger than social media. Research should be designed in a way that provides a holistic understanding of the relationship between youth well-being and technology and other factors (e.g., cultural and environmental factors that lead people to use technology and their use of technology, rather than focusing on Meta-specific product offerings.'' Thus, the impetus for their research appears to be advocacy or public relations rather than scientific discovery or public safety. Project MYST revealed two key findings. First, the study found that children who had adverse events in their lives were more vulnerable to ``potential problematic use'' (i.e., a gentle euphemism for social media addiction). Second, researchers found that parental supervision made no difference whatsoever to the addictive impact of Meta's social media platforms. The study found there was no relationship between a parent setting time limits or useful safety tools and whether their child became addicted. It is little surprise, then, that Meta buried the data and prohibited the study's results from ever seeing the light of day--until litigation in California State by private plaintiffs thrust the evidence into open court. YouTube documents also provided evidence of problematic usage by virtue of design. In 2015, YouTube launched Autoplay on desktop (followed by a launch in 2016 to app), which automatically and continuously played the next video--without any need for user action. Autoplay set to on be default because it generates more watch time. Autoplay watch time doubles at night. As one YouTube employee explained, ``This was the single most impactful launch in YouTube history, with +8 percent desktop watch time, +4 percent overall watch time increase.'' In YouTube's own documents, making the platform ``addictive'' was a core design feature. This was consistent with the aspiration of the same presentation of ``Building the world's most powerful and delightful video consumption experience.'' YouTube internally acknowledged the potential negative effects of digital videos in a 2018 presentation entitled ``Literature Review: Effects of watching digital videos and viewer well-being,'' which called out problematic Internet use in particular: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] The slide calls out problematic Internet use and states that it overlaps with ``addiction'' consistent with my belief that the entire continuum to the right of casual use (Figure 20) can be viewed as problematic. Furthermore, it highlights how ``autoplay,'' a key feature of YouTube, drives the ``just one more video effect.'' Slide 10 summarizes data from an internal survey of ``265 respondents'' that calls out the ``stickiness'' of the app and states that its interactivity and notifications ``causes users to feel that they must be aware of what is happening on the platform'' which ``keeps users on the platform longer.'' A 2018 YouTube presentation exploring the ``negative effects that video watching has on user's wellbeing'' included an express reference to addiction: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] TikTok's engagement algorithm, meanwhile, is frequently viewed as the most effective in the industry: ``more personalized,'' ``more accurate,'' and ``more diversified.'' The company emphasizes its effectiveness in its marketing presentations and touts that more than 53 percent of suggested videos are ``viewed.'' An internal memo citing the company's ``vulnerabilities'' reports: ``In a user survey of 2,300 users in February of 2020, when respondents were asked to give a score of 5 to indicate strength of agreement with the statement `I spend too much time on TikTok,' the average response was a 4.0.'' It goes on to say, ``some elements of persuasive design may be unique to TikTok; for instance, the fact that when you click the back button on your phone to leave the For You Feed/all, you get a prompt saying, `Tap again to exit,' which can be seen as increasing friction for users seeking to leave the app.'' And still later concedes, ``TikTok is particularly popular with younger users, who are seen as more vulnerable to online harms and the negative impacts of compulsive use.'' In a ``2021 TikTok for Good Business Plan and Vision'' document, the following is stated: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Notably, the plan acknowledges that ``addiction to technology is a ubiquitous problem.'' At the same time, comment [82] notes that ``addiction'' could be ``considered a very positive metric in our field.'' In other words, for all of its public posturing seeking to discredit or minimize the existence of compulsive or addictive use of screens, TikTok's internal documents acknowledge its existence and even allude to its ``value'' to the industry. TikTok's algorithms are widely considered the ``best in the industry'' because of its effectiveness at driving engagement. In pursuit of that end, TikTok made leaving the app more difficult than others. And for those times when people did succeed in leaving, TikTok refined its ``push'' approach to, among other things, get people back onto the company's platform. TikTok's ``push methodology,'' like every change they make to their platform, was subjected to A/B testing where rapid cycle experiments were performed comparing one version to another with users randomized to experience one or the other (i.e., no confounding) and with an eye to ensuring that core metrics were not adversely affected. Finally, Snap--the makers of Snapchat--innovated on the FOMO [``fear of missing out''] concept of maintaining ``streaks'' by developing the Snap Streaks feature, which it specifically designed to gamify its platform and drive usage and engagement. Streaks are built and maintained by two people ``snapping'' back and forth on consecutive days. The quality or content of the snap is irrelevant--it's all about ping-ponging back and forth consistently. Streaks were extraordinarily effective. Less than a year after they were launched, 22 percent of users had at least one streak, and the average highest streak value was 76.243. Snap Streaks were also particularly popular with younger users, as 47 percent of Streak users were under the age of 17.244. Of course, the true number was likely even higher, given how easy Snap--and every social media platform--has made it for children to lie about their age. Snap Streaks' importance to users was also tied to Snap Scores. While streak counts are private to the participants, every user of Snap has a numerical Snap Score displayed on their profile. The way that this score is generated is opaque, but broadly speaking it reflects how much they engage with the app and how much engagement they receive from others, including Streaks. For users, high Snap Scores are a way to ``show off their popularity.'' Snap's research found that ``users mainly used Streaks as a social status measure (to increase Snap Scores, have as many Streaks as possible) rather than a communication tool.'' This is consistent with Snap's conclusion that once users knew how to increase their score, it ``may make user care about (if not become addictive to) their [] score.'' However, Snap saw this as a positive feature of Snap Scores because it would ``encourage [users] to produce and consume more snaps.'' But at the same time that Snap was chasing the engagement increases offered by streaks, worries about the addictive effect of Streaks on children were growing. For example, focus groups results forwarded by the Senior Director of Corporate Communications and Public Affair to the Vice President of Global Public Policy found the following: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Another employee put it even more bluntly: ``we seem to have tapped into some mass psychosis where 17 million people must keep the streaks going.'' A 2017 study of Snap ``Power Users,'' commissioned by Snap to better understand features that drove usage, found that ``[f]or some, streaks have become a ``compulsive behavior'' and that they feel they are ``in too deep.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Similarly, an internal presentation on streaks tried to spin them as a positive tool for building friendships but nonetheless was forced to observe that Streaks can be ``really stressful'' and ``make[] it impossible to unplug for even a day.'' To better assess the harms of streaks, in 2018 Snap commissioned a survey of 790 users aged 13-24 (its core demographic): [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] In Snap's telling, this study showed that that only a minimal number of users found streaks stressful. If these charts seem familiar, it is because Snap presented this same conclusion to the United States Senate, writing that the study showed ``the majority of our community did not indicate Streaks were a significant source of stress--but six percent did.'' However, both the methodology and Snap's spin on the results were deeply flawed. Given Snap's millions of users, 6 percent still represents hundreds of thousands if not millions of users. And to get to 6 percent, Snap disregarded the ``moderate'' stress responses entirely. Snap also, without justification, combined those who found streaks a little stressful with those who found they caused no stress at all. A more accurate assessment of stress levels might be that 21 percent of users experience at least moderate stress because of snaps. A more holistic statement of the survey's results would be that 54 percent of streak users found streaks at least a little stressful. The other problem with sampling ``users'' to assess experiential stress is what in epidemiological terms is called the ``survivor effect.'' Many people who found streaks intolerably stressful would have left the platform or opted not to participate in a study about how stressful they are. In other words, this approach leads to a biased sample. Much like asking people still in a baseball stadium when the game has gone into 13 innings if games take too long will underestimate the true proportion of fans who think they do, asking regular users of an app how stressful it is underestimates the unease it induces. The study also depended on self-reporting from young people. But as Jennifer Stout, Snap's Global Vice President for Public Policy, pointed out, ``Kids like a lot of dumb things and parents are always trying to regulate their activities for their own safety!'' Snap's spin on the 2018 study is further undermined by the fact that Snap employees, including CEO Evan Spiegel, continued to express concern that streaks were addictive and harmful for users: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] A 2023 document containing suggested answers for an employee Q&A with Speigel is clear that Streaks ``can cause confusion and/or anxiety. We know this from the millions of support tickets we get every week from people asking us to restore a streak they accidentally lost.'' Users themselves make clear that they found streaks addictive. As the Q&A answer explained, users' frantic commitment to streaks could be tracked in part by tickets seeking to restore lost streaks. But the answer actually understated how desperate users were to have their streaks restored. By 2021, Snap was receiving an average of 400,000 streak restore requests each day, making up 95 percent of the total volume of customer service contacts. In some cases, individual users reach out directly to Mr. Spiegel to express the harms of streaks. Snap also explicitly embraced the idea that social rewards were key to keeping users coming back for more, conducting research that ``confirmed the hypothesis that Posters are motivated to post because audience feedback is their ROI reward.'' [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Indeed, Snap's research ``confirmed the causal relationship between receiving feedback (views and replies) and the poster propensity to post again.'' Following these conclusions, Snap's product team proposed tweaking ``Story reactions'' and replies to generate more engagement and posting. And just as feedback made users feel good, Snap's user research team found that not getting feedback was ``discouraging.'' Snap researchers found that 44 percent of Snapchat Story posters didn't receive any feedback on a given day leaving them more anxious and worse off than not having posted at all. Senior Snap employees were clear- eyed about the consequences of manipulating users' basic neurobiology and socioemotional responses. Responding to suggestions for ways to increase feedback rewards on Snapchat, Mr. Spiegel chimed in to say: [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Similarly, when reviewing proposed changes to the way that streaks operate, Stephen Collins, a Director of Public Policy, expressed a similar sentiment, observing that ``[r]ewards are known to drive compulsive/addictive behavior among some vulnerable groups.'' Of course, Collins's solution to this was not to reduce the role that rewards-maximizing play in Snap's design but just avoid using the word. Snap also used other forms of notifications to drive engagement. In 2023, Snap decided to implement an ``always on badge'' to get more users to try Spotlight. This badge was a red dot that was always present on the spotlight icon, visually drawing users to it. It would only go away once a visitor clicked on Spotlight and would reactivate every 60 minutes, regardless of whether there was any new content or other change to actually notify a user of. This one design change raised the daily active use of Spotlight from 53 million to 71 million, the single biggest increase in daily active use ever seen in an A/B test. Snap was not only made aware of the risks of these features, but they were also told to remove them by outside social media experts (SME's) they consulted. In sum, nearly every social media company adapts addictive and harmful design elements from their competitors. Separate and apart from the third-party content underpinning the platforms, these design elements insidiously work to keep users--typically children and adolescents--glued to the screens for longer and longer periods of time. Yet social media companies routinely turn to Section 230 to shield themselves from liability for these design choices. ______ Response to Written Questions Submitted by Hon. Maria Cantwell to Brad Carson AI and Section 230: There's bipartisan agreement that, despite the benefits of Section 230 in helping create the robust Internet ecosystem we have today, we have done too little to mitigate the harms of social media, especially to children. In 1996, Congress wanted to ensure that small websites wouldn't be sued for everything their users said. But Congress had no idea that Section 230 would shield all kinds of social media practices, like algorithms that hook kids. It's been thirty years, and we still don't have basic Federal protections for kids online or data privacy. Now, some want to take the model of Section 230--preemption or nullification of state laws without a Federal regulatory framework--and apply it to artificial intelligence. Question 1. What does the history of the last thirty years of Section 230 tell us about regulating AI today? Answer. The thirty-year history of Section 230 teaches a single, sobering lesson: broad immunity granted to an emerging industry during its infancy is extraordinarily difficult to correct once that industry matures. When Congress passed Section 230 in 1996, just 16 percent of Americans were online, Google did not yet exist, and Facebook was nearly a decade away. No one anticipated that a handful of platforms would come to control virtually all public communication, accumulate market capitalizations exceeding the GDPs of many countries, and engineer products deliberately designed to maximize engagement at the expense of user wellbeing. The immunity that followed those platforms as they matured was never seriously revisited, because the industry had grown powerful enough to resist correction and because the meta-law that Section 230 created determined who would govern AI rather than how strong that governance would be. The danger Congress now faces with artificial intelligence is structurally identical. Right now, there are voices urging Congress to freeze state laws without enacting comprehensive Federal regulation. In other words, those same voices are asking Congress to repeat the Section 230 model for a technology whose potential implications dwarf those of social media. The lesson of Section 230 is not merely that bad outcomes follow from a broad immunity; it is that the first major legal framework tends to become the last, because it allocates power to those who are most likely to resist future reform. As it relates to AI regulation, Congress should avoid creating the conditions for courts to improvise in a legal vacuum, so that it doesn't have to spend a generation debating whether the improvisation was correct. Question 2. Mr. Carson, AI agents now post, moderate, and curate online content. None of this existed when Congress wrote Section 230. How should the law apply when platforms generate and deliver content using AI, rather than simply hosting it? Answer. Section 230(c)(1) grants immunity to platforms for content ``provided by another information content provider.'' The statute contemplates a world of active users and passive platforms wherein the latter merely hosts what the former creates. That model does not describe how generative AI operates. When a user submits a prompt, the output is produced by a model that was designed, trained, fine-tuned, and deployed by the company that built it. The company made the choices that determined what the model would say. The output of a large language model is not the ``information content'' of ``another'' within the meaning of the statute. It is, in substantial measure, the product of the company itself. The same analysis applies to AI agents acting autonomously; they are not third-party users posting content, but rather automated expressions of design decisions made by their developers. Although I believe this textual answer is clear, I urge Congress to codify it. The lesson of Zeran v. America Online is precisely that textually plain answers do not always survive contact with courts under pressure to resolve novel cases quickly and in the absence of legislative guidance. Congress should clarify that Section 230 does not apply to generative AI outputs, and it should do so before lower courts, following the Zeran precedent, extend immunity beyond what the statute's text permits. Question 3. Mr. Carson, should Congress be cautious when providing immunity to a fast-changing technology? Answer. Yes, and the faster the technology changes, the more cautious Congress should be. Broad immunity provisions are difficult to repeal precisely because the industry they protect grows powerful enough to resist intervention. When Congress granted near-blanket immunity to the early internet, it could not have foreseen that a few platforms would come to control virtually all public discourse and command market capitalizations exceeding the GDPs of entire countries. The immunity that followed those platforms into dominance proved nearly impossible to reclaim. Creating broad immunity for an emerging industry should be recognized as a category error disguised as prudence; we should not permanently exempt an entity from legal responsibility because it appears harmless today. The better approach is to calibrate liability to capability and allow it to scale as the technology and its demonstrated harms develop. This is precisely what Section 230 failed to do, and it is what a responsible legal framework for artificial intelligence should accomplish. Flexible structures that assign liability standards to agencies, mandatory safety certification markets, and liability insurance calibrated to risk profiles are all examples of possible mechanisms that will grow with the technology rather than locking in protections derived from a moment wherein we have no clear conception of how this technology will evolve. Kids Online Safety: We heard at a hearing earlier this year on kids and screen time that AI poses new and complex threats to kids online. Kids are increasingly interfacing with chatbots that are shaping their social development and relationships. These AI companions, so-called virtual girlfriends or boyfriends, were not designed with kids or teens in mind, yet they are warping kids' development and their understanding and appreciation of healthy human relationships. Mr. Carson, your testimony highlighted the dangers that preemption of state AI laws poses to kids online. Yet Senate Republicans and the White House have made repeated attempts to preempt state AI laws, while there is no Federal standard in place. Given the risks new technologies like AI pose to kids' safety, we cannot ask states to stand down from protecting children--especially when Congress has yet to establish Federal standards. Question 1. Mr. Carson, how are states stepping up to protect kids from AI in the absence of a Federal standard? Answer. In the absence of Federal standards, state lawmakers have moved to fill the void and protect vulnerable constituencies in the AI era. Unfortunately, their efforts have been met with hostility from the White House. Earlier this year, Utah lawmakers introduced legislation that would require AI companies to publish child safety plans and to prevent chatbots from engaging in inappropriate conversations with minors. The White House then successfully lobbied the Utah legislature to kill that legislation. Florida this year considered an AI Bill of Rights that would affirm protections against explicit AI images, including those featuring minors, and that would provide parental controls on AI for minors. Again, the White House pressured state lawmakers, this time Florida, to drop the effort. What we're seeing from states is bipartisan interest in thoughtfully advancing AI safeguards, especially for minors, but those efforts have been repeatedly stymied by Administration officials who have prioritized a rules-free approach for the AI industry. Last month, over 50 state lawmakers from impacted states wrote to the White House, urging the Administration to stop its pressure campaign on state AI laws. The bottom line is that state lawmakers are working to respond to constituent concerns, but there's growing frustration with Federal efforts to preempt their legislation without a substantive replacement. Question 2. What risk does preemption of AI laws pose to children without meaningful Federal standards? Answer. Preemption of state AI laws in the absence of a meaningful Federal standard creates a legal vacuum that leaves children entirely unprotected. It removes states' authority to apply their tort law, consumer protection statutes, and civil liability regimes to AI systems that harm minors. It prevents injured families from legal redress before a Federal alternative is established. And it signals to courts that they are under no obligation to develop common law protections, while providing no alternative framework to fill that space. This is the Section 230 model, deliberately reproduced, and we should remember that children paid the highest price for that model's failures. The White House's recent pressure to block state-level protections in Utah and Florida, while offering no Federal substitute, illustrates the precise dynamic that Section 230 created: industry-friendly preemption dressed up as Federal uniformity. The cost of preemption must include a genuine Federal framework that replaces the protections it displaces. Congress cannot in good conscience instruct states to stand down from protecting children and then fail to enact the Federal protections that would justify asking them to do so. If preemption is to occur, it must be accompanied by legally enforceable Federal standards that give children at least equivalent protection to what the states would have provided and already are providing in many cases. Section 230 Liability Protection: Section 230 was adopted thirty years ago on a simple idea: platforms hosting user speech should not be treated as publishers of every post, and those that moderate harmful material should not be punished for doing so. That framework has been critical for startups, small businesses, and online communities that depend on the ability to host user content without being sued over every post. But thirty years of experience have also shown that the largest platforms used these protections very differently from how Congress intended. They built business models around algorithmic amplification, addictive design, and pervasive data collection. They hid behind Section 230 while their products harmed children and disrupted elections. As we consider Section 230 at thirty, I want to understand what has worked well and what hasn't. Specifically, what these protections have meant for free speech and for the smaller platforms that depend on them the most. But also how we hold tech platforms accountable for their harms. Question 1. Do you agree that large platforms need to be held accountable for their actions and the harms they are causing to society--particularly harms to children and data privacy? Answer. Yes, unequivocally. The thirty-year saga of Section 230 is a story about what happens when powerful platforms are permitted to externalize the costs of the harm they enable. All the families whose children were victimized by predators, parents whose teenagers experienced years of self-harm linked to engagement-optimization design choices, and individuals whose reputations were destroyed by algorithmically amplified falsehoods quickly discovered that the law offered them little or nothing. The platforms bore few of those costs. That asymmetry is not an accident of technology; it is the predictable consequence of a legal regime that removed any meaningful form of accountability as a check on corporate behavior. Large platforms must be held responsible both for preventing foreseeable harms and for compensating those harmed when prevention fails. It is worth noting that the harm enabled by large platforms extend well beyond children and other vulnerable populations, as serious as those are. Platforms have served as vectors for foreign influence operations, the coordination of political violence, and the rapid dissemination of content that has facilitated real-world harm including exploitation of infrastructure vulnerabilities. As AI dramatically amplifies the speed and scale at which content can be generated and targeted, the gap between the policies we need and the policies we have will widen unless Congress acts. The question is not whether platforms should be held accountable, but whether Congress will build the legal architecture to make that accountability real before AI makes the cost of inaction far greater than it has already been. ______ Response to Written Questions Submitted by Hon. Amy Klobuchar to Brad Carson State AI Preemption: You wrote that attempts to preempt state AI laws was ``essentially version 2.0 of Section 230, the liability shield that has allowed social media platforms to escape accountability for over a decade.'' Question 1. What lessons have we learned from giving social media platforms near-blanket immunity and how should we apply those lessons to AI? Answer. The thirty-year history of Section 230 teaches a single, sobering lesson: broad immunity granted to an emerging industry during its infancy is extraordinarily difficult to correct once that industry matures. When Congress passed Section 230 in 1996, just 16 percent of Americans were online, Google did not yet exist, and Facebook was nearly a decade away. No one anticipated that a handful of platforms would come to control virtually all public communication, accumulate market capitalizations exceeding the GDPs of many countries, and engineer products deliberately designed to maximize engagement at the expense of user wellbeing. The immunity that followed those platforms as they matured was never seriously revisited, because the industry had grown powerful enough to resist correction and because the meta-law that Section 230 created determined who would govern AI rather than how strong that governance would be. The danger Congress now faces with artificial intelligence is structurally identical. Right now, there are voices urging Congress to freeze state laws without enacting comprehensive Federal regulation. In other words, those same voices are asking Congress to repeat the Section 230 model for a technology whose potential implications dwarf those of social media. The lesson of Section 230 is not merely that bad outcomes follow from a broad immunity; it is that the first major legal framework tends to become the last, because it allocates power to those who are most likely to resist future reform. As it relates to AI regulation, Congress should avoid creating the conditions for courts to improvise in a legal vacuum, so that it doesn't have to spend a generation debating whether the improvisation was correct. Section 230 and AI: As on social media platforms, we've seen generative AI systems provide harmful content to children. Question 1. You argue that Section 230 should not apply to generative AI. What are the key differences between social media platforms and generative AI systems when it comes to the application of Section 230? Answer. The fundamental distinction lies in authorship. Section 230(c)(1) immunizes platforms for content ``provided by another information content provider''. In short, content that originates with a third-party user, not with the platform itself. A social media platform that hosts a post written by a user is, in the statute's terms, a passive conduit. A generative AI system is something categorically different; it produces its own outputs in response to user prompts, based on training data selected by the company, fine- tuning choices made by the company, and deployment parameters set by the company. The output is not the speech of ``another'' by any reasonable standard. It is, in substantial measure, the product of the developer's own design decisions. Section 230s passive-host immunity simply does not map onto an active content-generating system. A secondary, and more nuanced, question is whether AI companies qualify as ``interactive computer services,'' which is the threshold category that Section 230 protection requires. In my own view, there is a credible argument that a generative AI provider, unlike a traditional platform, does not provide a service through which users access content from others, but rather a system that itself generates content on demand. If that argument prevails in litigation, AI developers would fall outside Section 230s scope on both textual grounds simultaneously. Rather than leave the issue to serendipity, Congress should clarify the law before courts are forced to resolve these questions in a vacuum, producing precedents that may prove as difficult to correct as Zeran vs. AOL. ______ Response to Written Questions Submitted by Hon. Brian Schatz to Brad Carson AI Chatbots: Question 1. Do you believe that content produced by AI chatbots is protected by Section 230? Answer. No. Content produced by AI chatbots should not be protected by Section 230, and I do not believe that the statute's text, properly read, extends that protection. Section 230(c)(1) immunizes providers for content ``provided by another information content provider.'' A chatbot's output is not provided by ``another,'' but rather it is generated by the system that the company designed, trained, and deployed. The user supplies a prompt; the company's model supplies the response. The company is, in meaningful and legally significant measure, the author of that response. In my view, this is not an ``umpire call'' on a close textual question, but I urge Congress to codify the answer before courts, under pressure to resolve novel cases in a legal vacuum, extend immunity in ways that will result in an entire generation of children being harmed before we can correct it. Question 2. If AI companies can claim immunity from harmful AI- generated content, what consequences do you predict for users and platforms? Answer. The consequences would be both serious and predictable. AI companies shielded from liability for harmful outputs would face no legal incentive to invest in safety measures beyond what their commercial interests already demand. Victims of AI-generated harm, whether that harm takes the form of defamation, non-consensual intimate imagery, content that facilitates self-harm in vulnerable users, instructions enabling real-world violence, or outputs used to manipulate individuals in ways that damage their financial or physical wellbeing, would be denied access to legal remedies. The legal system would have no mechanism through which to send price signals to AI developers about the externalized costs of their systems' failures. And the absence of liability would remove one of the most powerful incentives for independent safety auditing, since companies would have little legal reason to surface risks they are not required, by law, to remedy. More generally, broad immunity would entrench a two-tier system: well-resourced actors who can absorb the harms AI inflicts on them, and everyone else, who cannot. The families, teenagers, and individual users who have already suffered under Section 230s regime are the same populations most vulnerable to AI's harms. Congress has a choice between mechanisms that establish accountability proportional to model capability, and immunities that are inversely proportional to foresight. The story of Section 230 makes clear which choice ages better. [all]