Full text of the official published hearing record. Extracted from the source document — verify against the official record for citation.
- INTELLIGENT NETWORKS: POWERING ARTIFICIAL INTELLIGENCE AND TRANSFORMING COMMUNICATIONS [Senate Hearing 119-535] [From the U.S. Government Publishing Office] S. Hrg. 119-535 INTELLIGENT NETWORKS: POWERING ARTIFICIAL INTELLIGENCE AND TRANSFORMING COMMUNICATIONS ======================================================================= HEARING BEFORE THE SUBCOMMITTEE ON TELECOMMUNICATIONS AND MEDIA OF THE COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION UNITED STATES SENATE ONE HUNDRED NINETEENTH CONGRESS SECOND SESSION __________ JULY 30, 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 65-509 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 ------ SUBCOMMITTEE ON TELECOMMUNICATIONS AND MEDIA DEB FISCHER, Nebraska, Chair BEN RAY LUJAN, New Mexico, Ranking JOHN THUNE, South Dakota AMY KLOBUCHAR, Minnesota ROGER WICKER, Mississippi BRIAN SCHATZ, Hawaii JERRY MORAN, Kansas EDWARD MARKEY, Massachusetts DAN SULLIVAN, Alaska GARY PETERS, Michigan MARSHA BLACKBURN, Tennessee TAMMY BALDWIN, Wisconsin TODD YOUNG, Indiana TAMMY DUCKWORTH, Illinois TED BUDD, North Carolina JACKY ROSEN, Nevada ERIC SCHMITT, Missouri JOHN HICKENLOOPER, Colorado JOHN CURTIS, Utah JOHN FETTERMAN, Pennsylvania BERNIE MORENO, Ohio ANDY KIM, New Jersey TIM SHEEHY, Montana LISA BLUNT ROCHESTER, Delaware SHELLEY MOORE CAPITO, West Virginia CYNTHIA LUMMIS, Wyoming C O N T E N T S ---------- Page Hearing held on July 30, 2026.................................... 1 Statement of Senator Fischer..................................... 1 Letter dated July 30, 2026 to Senator Deb Fischer and Senator Ben Ray Lujan from David A. Wright, Policy Director, Spectrum for the Future.................................... 3 Letter dated July 28, 2026 to Hon. Deb Fischer and Hon. Ben Ray Lujan from Sharon Wilson Geno, President, National Multifamily Housing Council; Bob Pinnegar, President and Chief Executive Officer, National Apartment Association; and Kevin Donnelly, Executive Director and Chief Advocacy Officer, Real Estate Technology & Transformation Center.... 5 Letter dated July 30, 2026 to Senator Deb Fischer and Senator Ben Ray Lujan from Mary L Brown, Executive Director, WifiForward................................................ 7 Letter dated July 30, 2026 to Sen. Deb Fischer and Sen. Ben Ray Lujan from ConnectAI................................... 8 Letter dated July 29, 2026 to Sen. Deb Fischer and Sen. Ben Ray Lujan from Shawn M. Vasell, Vice President, Corporate Affairs, Hewlett Packard Enterprise........................ 10 Letter dated July 30, 2026 to Hon. Deb Fischer and Hon. Ben Ray Lujan from John Puskar, Chief Executive Officer, American-Made 5G Coalition................................. 11 Letter to Hon. Deb Fischer, Hon. Ben Ray Lujan, Hon. Ted Cruz and Hon. Maria Cantwell from Gary Bolton, President and CEO, Fiber Broadband Association........................... 12 Statement of Senator Lujan....................................... 21 Prepared statement........................................... 22 Statement of Senator Klobuchar................................... 45 Statement of Senator Capito...................................... 47 Statement of Senator Rosen....................................... 49 Statement of Senator Cruz........................................ 50 Statement of Senator Blunt Rochester............................. 53 Statement of Senator Young....................................... 54 Statement of Senator Cantwell.................................... 56 Witnesses Jonathan Spalter, President and Chief Executive Officer, USTelecom--The Broadband Association........................... 14 Prepared statement........................................... 15 Dan Watermeier, Commissioner, Nebraska Public Service Commission. 23 Prepared statement........................................... 24 Bob Everson, Chief Architect, Provider Mobility, Cisco Systems, Inc............................................................ 25 Prepared statement........................................... 27 Asad Ramzanali, Director of Artificial Intelligence and Technology Policy, Vanderbilt Policy Accelerator, Vanderbilt University..................................................... 32 Prepared statement........................................... 33 Appendix Response to written questions submitted to Jonathan Spalter by: Hon. Maria Cantwell.......................................... 63 Hon. John Hickenlooper....................................... 64 Response to written questions submitted to Bob Everson by: Hon. Edward Markey........................................... 65 Hon. John Hickenlooper....................................... 65 Response to written questions submitted to Asad Ramzanali by: Hon. Maria Cantwell.......................................... 66 Hon. John Hickenlooper....................................... 69 Hon. Brian Schatz............................................ 70 INTELLIGENT NETWORKS: POWERING ARTIFICIAL INTELLIGENCE AND TRANSFORMING COMMUNICATIONS ---------- THURSDAY, JULY 30, 2026 U.S. Senate, Subcommittee on Telecommunications and Media, Committee on Commerce, Science, and Transportation, Washington, DC. The Subcommittee met, pursuant to notice, at 10:03 a.m., in room SR-253, Russell Senate Office Building, Hon. Deb Fischer, Chair of the Subcommittee, presiding. Present: Senators Fischer [presiding], Cruz, Young, Capito, Lujan, Cantwell, Klobuchar, Rosen, Hickenlooper, Fetterman, and Blunt Rochester. OPENING STATEMENT OF HON. DEB FISCHER, U.S. SENATOR FROM NEBRASKA Senator Fischer. Good morning to the members of this subcommittee, and to our witnesses. Thank you all for being here. This hearing will explore the relationship between telecommunications infrastructure and artificial intelligence. That relationship moves in both directions. AI requires low latency and high bandwidth networks, but AI has real potential to make networks more efficient and secure. This hearing is not intended to be a referendum on the merits of artificial intelligence. Around the world, however, our allies and adversaries are jockeying for AI leadership. To position the United States for ongoing economic and security leadership, the U.S. needs AI-supported and AI- enhanced networks. AI is increasingly a part of Americans' daily lives. For example, in 2024, nearly three-quarters of enterprises used AI for at least one business function. In 2025, more than half of Americans reported using generative AI. These trends have been accompanied by a surge in demand across networks, and AI adoption is only expected to grow. Most enterprises are still scaling their uses of AI, and many individual AI uses are for basic search functions. As AI use, and downstream applications of AI continue to grow, demand for reliable, high-speed, high-capacity networks will also grow. Today, we will explore how the rapid adoption of AI has impacted existing network infrastructure. This includes the new demand that AI has added to networks. We will explore how widespread AI use has forced networks to evolve, requiring more capacity and more complex designs so that AI can run efficiently on those networks. We will consider how government providers and other industries are responding to that demand. Private companies have invested hundreds of billions of dollars in network deployment in recent years. Various Federal broadband programs have also provided billions to support targeted network deployment and maintenance throughout the country. We will review obstacles to AI-supportive networks, opaque regulations and lack of coordination should not get in the way of network deployment, of network development, and the many downstream applications it enables. And finally, we will consider how AI can be used to improve our networks. Zooming out from the technical questions, we must also ask a more fundamental question: Why must our networks be prepared for AI? First, America has a geostrategic imperative to make sure our networks are the best in the world. Now more than ever, we cannot fall behind in developing the most resilient and secure networks. Second, to reap the full economic benefits from AI and other rapidly emerging technologies, many of them are promising, and we need reliable connectivity for individuals and enterprises all over the country. This is not connectivity for the sake of just checking a box, or hitting a milestone. Nebraska's hardest-to-reach communities include farms and ranches that are at the leading edge of connectivity-based ag solutions. Rural hospitals depend on resilient and secure communication networks to use state-of-the-art medical technologies, and reach patients in remote areas. Manufacturers rely on the Internet of Things, on robotics, and cloud-based systems to streamline internal processes and improve quality control. As AI adoption increases, so will the demand for reliable and resilient communications infrastructure. Networks are the backbone, along with the enormous amounts of data associated with AI are transmitted. I look forward to discussing both how networks are adapting to respond to AI and how AI is being used in networks to proactively plan for the future. Thank you. Before we begin, I would ask unanimous consent that letters from a broad range of organizations describing the need for AI- ready communication networks, be entered into the record. We have a number of them. Without objection, so ordered. [The information referred to follows:] Spectrum for the Future July 30, 2026 Senator Deb Fischer, Chairwoman, Subcommittee on Telecommunications and Media, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Senator Ben Lujan, Ranking Member, Subcommittee on Telecommunications and Media, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Re: Hearing on ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications'' before the Committee, July 30, 2026 Dear Chairwoman Fischer and Ranking Member Lujan, Spectrum for the Future is a diverse coalition of innovators, anchor institutions, and technology companies using shared licensed spectrum such as the 3.5 GHz CBRS to further America's wireless technology leadership. As the Committee examines how the United States can maintain its leadership in artificial intelligence (``AI''), Spectrum for the Future encourages Members of the Committee to advance policies that preserve the Citizens Broadband Radio Service (``CBRS''), including maintaining current power levels, and expanding access to spectrum for the Private 5G deployments that enable AI. We applaud the Committee's efforts in holding this hearing and write to amplify the hearing record in support of extending American leadership in AI, particularly as it relates to 1) the use of private mobile networks as a critical communications layer for the implementation of AI by American factories, ports (air and sea), farms, warehouses, and industries broadly, and 2) AI-based innovations in the dynamic management of radio frequency (``RF'') spectrum. As a direct result of CBRS innovations, the United States is the global leader in both private mobile networks and dynamic spectrum management. We urge the Committee to leverage and support this leadership as AI becomes increasingly critical to the American economy, national defense, and technological innovation. CBRS Private Networks enable AI Deployment American industry, businesses, and the military are implementing advanced AI services with CBRS-based private mobile networks as the foundation. According to an ongoing series of reports from the Global mobile Suppliers Association (``GSA''), the United States leads the world in private mobile networks due to the availability of CBRS spectrum.\1\ The GSA report further notes that CBRS is the leading spectrum band globally for the deployment of private mobile networks. CBRS supports both Private LTE (``pLTE'') and Private 5G (``p5G'') today and will support Private 6G (``p6G'') when it comes to market. --------------------------------------------------------------------------- \1\ Private Mobile Networks June 2026, Global mobile Suppliers Association, June 2026 --------------------------------------------------------------------------- AI services enabled by CBRS private network communications include Digital Twins, Computer Vision, Automated Monitoring, and Route Optimization. A few examples: John Deere uses AI to model Digital Twins of its manufacturing facilities utilizing CBRS-connected sensors, optimizing the layout of its production lines and proactively scheduling machine maintenance in advance of failures.\2\ --------------------------------------------------------------------------- \2\ John Deere's private 5G network is 'Never Idle', LightReading, April 1, 2025 The U.S. Navy has harnessed AI Computer Vision to analyze the video from CBRS-connected high-resolution rapid imaging devices of P-8 Poseidon aircraft engines, reducing inspection times by almost 95 percent, from 4 hours to 15 minutes, greatly --------------------------------------------------------------------------- improving operational availability. Miami International Airport utilized its CBRS private network to deploy a next generation perimeter intrusion detection system (PIDS), a cutting-edge AI security technology that identifies and responds to potential airport perimeter breaches in real-time.\3\ --------------------------------------------------------------------------- \3\ https://news.miami-airport.com/mia-showcases-new-perimeter- intrusion-detection-system/ The U.S. Marine Corps is not only ``first to fight'', but was also the first service branch to obtain an Authority to Operate (``ATO'') for its Private 5G network covering nearly 1 Million ft\2\ of Smart Warehouse space at its logistics base in Albany, GA. Among other uses, AI is utilized to monitor and optimize the routing of CBRS-connected forklifts and robots in the Automated Storage and Retrieval System (AS/RS) and inventory management process.\4\ --------------------------------------------------------------------------- \4\ https://www.nationalspectrumconsortium.org/api/documents/ 67a272176b75cbad79eeebb5/data These are just a few leading examples of the AI innovation that CBRS is enabling across America. AI and Dynamic Spectrum Management Dynamic Spectrum Management (aka Dynamic Spectrum Sharing) represents a uniquely American innovation, combining advanced digital spectrum coordination with equipment designed to optimize the use of spectrum by all users. This approach harnesses the best of American ingenuity, leveraging our leadership in AI and machine learning to dynamically manage spectrum resources in real time. Key benefits of dynamic spectrum management include: Enabling Federal and private sector spectrum sharing without displacing incumbent users, enabling America's national security and domestic economic interests to progress in tandem. Accelerating spectrum access for new commercial services without clearing incumbents. Ensuring spectrum policies remain innovative, secure, and adaptable to evolving technology. Spurring American job creation across a range of industries thru private networks innovation. Superior to ``command and control'' approaches such as those used by the Chinese Communist Party. To underscore this point, in March the Congressional U.S.-China Economic and Security Review Commission recommended that Congress ``Increase funding for foundational research on advanced technologies such as cognitive radio and radar, dynamic approaches to spectrum allocation, and spectrum sharing.\5\ --------------------------------------------------------------------------- \5\ China's Management of Electromagnetic Spectrum Resources, U.S.- China Economic AND Security Review Commission, March 12, 2026 With Dynamic Spectrum Management, the central cloud services coordinating access to spectrum (e.g., the CBRS Spectrum Access System (``SAS'') databases) have comprehensive awareness of what spectrum is or is not available. This makes these spectrum coordinating services ``fertile soil'' for the development of even more advanced AI capabilities to further optimize spectrum availability and wireless operations. As one example, Federated Wireless has released its Spectrum AI platform for CBRS, whose capabilities far exceed the regulatory minimum requirements, and which can increase overall network capacity as much as 5 times.\6\ --------------------------------------------------------------------------- \6\ https://federatedwireless.ai/products/spectrum-ai/ --------------------------------------------------------------------------- AI-driven Dynamic Spectrum Management is a testament to the United States' leadership, reinforcing our position as the global leader in both AI and wireless innovation. By prioritizing the extension of this leadership, we will ensure that America's communications infrastructure continues to serve the needs of its citizens, businesses, and government, keeping our Nation secure and competitive in the 21st century. Spectrum for the Future looks forward to working with the Committee on its important initiatives around AI and communications. We welcome the opportunity to provide additional information to the Committee on this important issue. Very Respectfully, David A. Wright, Policy Director, Spectrum for the Future. spectrumfuture.com ______ Real Estate Technology and Transformation Center July 28, 2026 Hon. Deb Fischer, Chairwoman, Senate Commerce, Science, and Transportation, Subcommittee on Telecommunications and Media, Washington, DC. Hon. Ben Ray Lujan, Ranking Member, Senate Commerce, Science, and Transportation, Subcommittee on Telecommunications and Media, Washington, DC. Dear Chairwoman Fischer and Ranking Member Lujan: On behalf of the Real Estate Technology and Transformation Center (RETTC), the National Multifamily Housing Council (NMHC) and the National Apartment Association (NAA), we write to thank you for convening the upcoming subcommittee hearing, ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications,'' and to share insight into how rental housing providers are leveraging artificial intelligence to expand affordable broadband access for the more than 44 million Americans who call rental housing home. RETTC, NMHC, and NAA provide a single voice for rental housing developers, owners and operators and the technology suppliers that are driving innovation and helping address our Nation's long-term housing challenges. One-third of all Americans call a rental property home, where technology solutions are increasingly leveraged to modernize property operations, improve housing affordability and enhance the resident experience. Broadband is no longer a luxury for renters. It is essential infrastructure for education, employment, healthcare, civic engagement and daily life. As the subcommittee examines how artificial intelligence can improve communications networks, we urge you to consider the distinct role of rental housing in both efforts. Bulk Billing and Managed Wi-Fi Are Proven Tools for Closing the Digital Divide Rental housing communities face particular challenges in delivering reliable, high-speed Internet access, especially in older buildings, high-density urban areas, and historically underserved communities. Bulk billing and managed Wi-Fi arrangements have emerged as powerful tools for overcoming these barriers. Bulk billing and managed Wi-Fi agreements eliminate many of the practical barriers consumers face when attempting to access broadband, including equipment rental fees, credit checks, security deposits and complex sign-up processes. Costs are typically incorporated into rent or charged as a transparent technology or amenity fee disclosed throughout the leasing process. As a result, these arrangements have been particularly effective in expanding broadband access in affordable, senior, student, veteran, and workforce housing, where ease of access and predictable, pro-consumer pricing are critical. Bulk arrangements also drive infrastructure investment. The known rate of return guaranteed in a bulk contract justifies the substantial initial capital expenditure needed to build out a modern broadband network. This is particularly true in existing affordable housing stock where low take rates and high turnover under a traditional retail subscriber model can make a project unprofitable for the Internet service provider. This has fostered a competitive market that allows smaller, independent providers to participate and encourages larger providers to serve markets they might otherwise forgo. Despite this progress, a meaningful segment of renters still requires targeted support to overcome affordability barriers to broadband adoption. Federal efforts have too often failed to reflect how broadband is delivered in modern rental housing. Any Federal framework for broadband planning, mapping or affordability assistance should be technology-neutral and should ensure that support reaches renters equitably regardless of whether connectivity is provided through individual retail subscriptions or community-wide solutions. AI Is Strengthening Network Security and Resilience in Rental Housing Rental housing communities are increasingly deploying smart home and IoT technologies ranging from access controls and locks to lighting, thermostats, and leak detection sensors. These systems depend on always-on, property-wide connectivity, and AI is helping housing providers and their technology partners secure and sustain them through: Real-time threat detection: isolating unusual behavior that could indicate network breaches or device vulnerabilities. Maintenance prediction: identifying network or equipment failures before they affect residents and minimizing downtime. Predictive network optimization: managing bandwidth to sustain performance during high-traffic periods. Smart network design: modeling infrastructure needs to guide cost-effective deployment that avoids overbuilding while ensuring equitable access. These capabilities depend on the coherent network architecture that bulk and managed Wi-Fi arrangements make possible. Fragmented service degrades performance, complicates security monitoring and impairs property operations that rely on modern connected systems. As policymakers examine how AI can strengthen the resilience of communications networks, we encourage greater attention on multifamily communities, where millions of Americans actually connect. Conclusion As Congress explores the future of AI policy and connectivity infrastructure, we urge you to: Recognize bulk billing and managed Wi-Fi as key affordability tools and ensure they remain fully compatible with any new or restructured broadband assistance, planning, or mapping framework. Preserve the existing partnership model that enables rental housing operators and communications providers to enter into agreements negotiated under free market conditions, encouraging competition, higher customer service standards and better, faster, more reliable service than is typically available in the broader retail market. Ensure that Federal examination of AI and network security accounts for multifamily communities, including the IoT and smart building systems that operate on property-wide networks. Prevent duplicative and fragmented AI regulations at the state and local levels, which risks stifling innovation and increasing compliance costs in ways that undermine the benefits these technologies offer renters and housing providers alike. Engage with housing providers when shaping AI, network security and broadband infrastructure policy, ensuring solutions account for the challenges, and opportunities across the rental housing landscape. RETTC, NMHC, and NAA stand ready to serve as a resource to the subcommittee as it continues this important work. We welcome the opportunity to share data, practical insights, and real-world experience from the rental housing sector to help inform durable, technology-neutral policies that expand broadband access and affordability for renters across the country. Thank you again for your leadership in exploring how secure, modern connectivity infrastructure can support a more connected future across rental housing. Sincerely, Sharon Wilson Geno, President, National Multifamily Housing Council. Bob Pinnegar, President and Chief Executive Officer, National Apartment Association. Kevin Donnelly Executive Director and Chief Advocacy Officer, Real Estate Technology & Transformation Center. ______ WifiForward July 30, 2026 Sen. Deb Fischer, Chair, Telecommunications and Media Subcommittee, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Senator Ben Lujan, Ranking Member, Telecommunications and Media Subcommittee, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Re: Hearing on ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications'' Dear Chair Fischer and Ranking Member Lujan, As the Committee examines the communications infrastructure needed to support America's AI leadership, WifiForward urges you to recognize the foundational role Wi-Fi and unlicensed spectrum play in the deployment of AI. In particular, we urge the Committee to support the global adoption of the Wi-Fi in the 6 GHz band and to allow Wi-Fi services to operate in the lower 7 GHz band. For the United States to lead in AI, every American needs robust connectivity, so that they can build and use AI tools. Wi-Fi serves as the lynchpin of connectivity for Americans nationwide, carrying significantly more consumer and enterprise data traffic than any other wireless technology. For example, 80-90 percent of all mobile traffic travels over Wi-Fi--not over cellular networks. Congress, the Federal Communications Commission (FCC), and the Executive Branch created the world's most AI-ready spectrum by opening the full 1200 megahertz in the 6 GHz band. That foresight launched a wave of innovation needed to support our Nation's AI ambitions.\1\ As AI begins to reshape network traffic, the 6 GHz decision has proven to be remarkably forward-looking. Moreover, the U.S. approach stands in stark contrast to China's emphasis on cellular infrastructure to support AI deployment while ignoring Wi-Fi, providing a strategic advantage for the U.S. --------------------------------------------------------------------------- \1\ FCC, Unlicensed Use of the 6 GHz Band, FCC 20-51 (Apr. 23, 2020). --------------------------------------------------------------------------- A recent report from Cisco Systems explains AI is not only changing both the volume and character of network traffic, but also the way network traffic growth behaves, creating new demands on infrastructure and in particular, Wi-Fi.\2\ These changes should prompt policymakers to direct limited spectrum resources support the technologies that will enable America's AI leadership. --------------------------------------------------------------------------- \2\ Cisco Systems, Inc., ``AI Impact on Wide Area Networks'' (2026) at https://www.cisco.com/c/dam/en/us/solutions/collateral/artificial- intelligence/mass-scale-infrastructure/ai-network-traffic-report.pdf. --------------------------------------------------------------------------- AI's center of gravity is both local and indoors. Data flows generated by AI software on a laptop, AI-connected cameras in a factory, an AI diagnostic system in a hospital, or an AI-powered robot moving through a warehouse all begin on local networks--and all rely heavily on Wi-Fi to move the AI traffic. Unlike wide-area mobile networks, which connect users across geographic areas, Wi-Fi is designed for the high-capacity, dense environments where many AI applications will operate. Wireline networks remain essential, but Wi- Fi provides the flexibility needed at the edge for scale and rapid deployment. Cisco's analysis lays out several important distinctions: 1. Unlike humans, AI systems can operate continuously, naturally resulting in more network traffic, with agentic AI generating up to 450 percent more traffic than a person performing the task manually,\3\ a trend that highlights how the traffic burdens on Wi-Fi networks are likely to grow significantly in the coming years. --------------------------------------------------------------------------- \3\ Antich, Javier, and Gurudatt Shenoy. ``The Impact of AI on Wide Area Network Traffic: We Need to Talk.'' Cisco Blogs, 21 May 2026, https://blogs.cisco.com/sp/the-impact-of-ai-on-wide-area-network- traffic-we-need-to-talk. 2. AI data flows are twice as long as non-AI traffic, resulting in the consumption of more spectrum resources. For wireless networks, duration can be just as important as volume. Longer flows mean greater airtime occupancy, more contention among --------------------------------------------------------------------------- devices, and increased pressure on shared spectrum resources. 3. AI traffic is increasingly bidirectional, with a growing emphasis on traffic moving from the user to the cloud--a reversal from download-optimized data traffic flows typically seen to date. That shift matters because Wi-Fi is especially well suited to handle more symmetrical traffic: unlike cellular networks, which are typically engineered around download-heavy consumer use. 4. AI traffic is increasingly autonomous, not pausing for human interaction, enabling AI to work without breaks or sleep, and generating network activity around the clock. This increased load falls predominantly on Wi-Fi because AI applications are being deployed where Wi-Fi networking dominates: inside homes, workplaces, factories, healthcare facilities, and other dense device environments. In addition, the vast majority of AI- connected devices operate on Wi-Fi and lack cellular connectivity. For those adopting the newer Wi-Fi 6 GHz-enabled technologies, there is good news. The Wi-Fi ecosystem is responding through technologies such as Wi-Fi 7 and the upcoming Wi-Fi 8, which use wider channels, lower latency, and more efficient coordination to accommodate AI workloads. Those technological advances, however, depend on sufficient spectrum. Even the most efficient protocol can become congested if traffic outpaces the available spectrum bandwidth available to Wi-Fi networks. Technology alone cannot meet increasing AI demands if spectrum policy fails to keep pace. As AI traffic continues to grow, policymakers must revisit their assumptions about future spectrum requirements for different technologies. If future traffic is more persistent, more autonomous, more uplink-intensive, more device-dense, and increasingly driven by software agents rather than people, then traditional forecasting models may no longer provide a complete picture of future capacity requirements for technologies like Wi-Fi that are poised to carry the overwhelming share of this incremental wireless traffic. At a minimum, policymakers need to preserve the full 6 GHz band for Wi-Fi. Beyond 6 GHz, spectrum planners should also be thinking about the next set of frequencies that would accommodate Wi-Fi's growing demand.\4\ For example, as the Executive Branch studies the feasibility of allowing high power cellular systems into segments of the 7 GHz band pursuant to the Working Families Tax Cuts Act, Pub. L. No. 119-21 (2025), we urge policymakers to consider frameworks that also allow Wi- Fi co-existence in the band. A balanced approach will be critical to supporting the next generation of AI connectivity while enabling innovation across the broader wireless ecosystem and delivering true win-win outcomes. --------------------------------------------------------------------------- \4\ Bill Davenport, Unlocking the Future: Why 6 GHz Wi-Fi Is the Foundation for the AI Era, Cisco Blogs (Apr. 2, 2026). --------------------------------------------------------------------------- Respectfully submitted, Mary L Brown, Executive Director, WifiForward. ______ ConnectAI July 30, 2026 Sen. Deb Fischer, Chair, Senate Committee on Commerce, Science, and Transportation, Subcommittee on Telecommunications and Media, Washington, DC. Sen. Ben Ray Lujan, Ranking Member, Senate Committee on Commerce, Science, and Transportation, Subcommittee on Telecommunications and Media, Washington, DC. Re: Hearing on ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications'' Before the Committee Dear Chair Fischer and Ranking Member Lujan, As the Committee examines the communications infrastructure needed to support America's leadership in Artificial Intelligence (AI), ConnectAI urges the Committee to take up legislation that ensures permitting keeps pace with modern communications infrastructure and streamlines the deployment of fiber optic networks just as Congress did for the legacy communications networks of years past. Otherwise, there is a risk of fiber connectivity becoming the next AI infrastructure bottleneck. In particular, policymakers should ensure that Federal communications permitting policy reflects how modern networks are built, including the essential support facilities that allow long haul fiber networks, the foundation of our entire communications system, to operate. For the United States to lead in AI, public policy must support the deployment of capable and sufficient connectivity, delivered where and when it is needed, so Americans can both build and utilize AI technologies. ConnectAI is a coalition of companies investing in the physical infrastructure required to deliver fast, reliable connectivity across the United States. Our members include leaders in fiber networks, wireline facilities, and the supporting structures that form the backbone of the digital economy. We are committed to expediting the deployment of terrestrial fiber because physical infrastructure is the prerequisite for innovation and an essential component of maintaining America's global leadership in AI. Much of the discussion surrounding AI infrastructure focuses on data centers, compute capacity, and energy resources. Equally important, however, are the fiber networks that connect these assets. Every AI workload depends on high-capacity, low-latency communications infrastructure to move data between users, cloud providers, data centers, and edge computing facilities. As AI applications become more sophisticated, more autonomous, and increasingly distributed, the demands placed on the Nation's communications backbone will continue to grow. Massive upgrades to the network are needed quickly to keep up with the pace of this demand. Fiber networks carry the traffic behind cloud computing, data centers, broadband networks, and wireless backhaul. These networks rely on facilities known as in-line amplifier facilities placed approximately every 50 miles to regenerate optical signals and maintain reliable communications across long distances. Without this infrastructure, the communications networks supporting AI and advanced computing could not function at scale. Because these facilities are deployed across nearly every state and hundred, if not thousands, of local jurisdictions, inconsistent siting and permitting requirements can delay construction of the infrastructure that powers America's AI economy. Permitting should provide the same regulatory certainty for these essential facilities that it has long provided for other traditional communications infrastructure. The siting and permitting of inline amplifier facilities, known as ``fiber huts'', which house the optical equipment necessary for long-haul routes, must be standardized and streamlined if deployments are to keep pace with demand. Fiber investment produces benefits that extend far beyond AI. Fiber supports rural broadband deployment, enables investment in new data centers, strengthens economic competitiveness, and creates lasting local economic benefits. Fiber infrastructure also provides greater flexibility in the placement of computing resources, helping reduce strain on existing power infrastructure while supporting growing demand for digital services. Most importantly, investment in domestic communications infrastructure ensures that the technologies, jobs, and innovation associated with AI remain in the United States. Technology leadership depends on infrastructure leadership. As the Committee considers policies to support America's AI future, ConnectAI encourages the Committee to modernize the permitting process, so it reflects the demands of today's communications infrastructure and streamlines deployment of the fiber optic networks that power intelligent networks. Fiber infrastructure plays an indispensable role in connecting the Nation's digital economy, and ensuring its timely deployment will strengthen America's leadership in artificial intelligence and beyond. Respectfully submitted, ConnectAI. ______ Hewlett Packard Enterprise July 29, 2026 Senator Deb Fischer, Chair, Telecommunications and Media Subcommittee, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Senator Ben Ray Lujan, Ranking Member, Telecommunications and Media Subcommittee, Senate Committee on Commerce, Science, and Transportation, Washington, DC. Re: Hearing Entitled ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications'' Dear Chair Fischer and Ranking Member Lujan, On behalf of Hewlett Packard Enterprise (HPE), thank you for convening this hearing on the role artificial intelligence is playing in the future of America's communications networks. This is one of the defining infrastructure questions of the decade, and HPE welcomes the opportunity to share how our technology is already helping solve it. The global race to next-generation connectivity will not be won through any single technology. It will be won by fielding a complete, trusted stack: advanced radio access technology, edge compute, secure data centers, and Al-native network management, all working together. HPE is one of the few U.S.-headquartered companies building solutions across that entire stack, from Juniper's carrier-grade routing to Aruba's Al-driven wireless, private wireless, and edge compute. We believe that the development of an American full-stack capability is what will determine whether the United States or its competitors set the terms for the next generation of global communications. HPE is leading the way in embedding Al into communications networks, from deploying real-time Al inferencing at the network edge to advancing Al-native network operations. We are currently running U.S. trials using HPE Proliant servers deployed directly at network edge sites, bringing Al capabilities closer to users and data sources. HPE is also conducting testing to apply machine learning to some of wireless communication's most persistent challenges. These capabilities are already moving from trials to real-world impact. In May 2026, HPE announced new self-driving network capabilities across HPE Mist and HPE Aruba Central that enable networks to detect, diagnose, and resolve issues in real time without human intervention. Customers have deployed HPE Self-driving Network capabilities to reduce service desk tickets by approximately 75 percent while bringing management of devices in-house-demonstrating how autonomous networking can improve reliability, reduce operational burdens, and support mission-critical services at scale. These efforts span three areas where Al is already improving how networks perform: Network reliability and safety: By training deep neural networks to predict user movement and dynamically adjust to changing signal conditions, we are improving indoor localization and anomaly detection capabilities that will improve public safety applications, such as helping emergency responders pinpoint a person's exact location during a medical emergency. Energy efficiency: Al-driven models that place antennas closer to users and predict mobility patterns more accurately are reducing the energy cost of signal transmission and handover, supporting the kind of reliable network buildout the country will need as data demand accelerates. Preparation for 6G: GPU-accelerated training of adaptive algorithms is laying the technical foundation for the shift from SG to 6G, addressing not only higher data rates but emerging requirements like precision location and sensing. The technical choices being made today on open interfaces, disaggregated network functions, and Al-native management will shape global network architecture for the next two decades. Other nations are investing heavily to influence those outcomes. American leadership in Al-enabled networking depends not just on innovation at companies like HPE, but on a policy environment spanning spectrum policy, standards engagement, and support for trusted domestic vendors that lets innovation reach the market to compete globally. HPE is committed to supporting the researchers, carriers, and infrastructure providers building the next generation of American communications capabilities, and we stand ready to serve as a resource to the Committee as it considers policy in this space. Thank you again for your leadership on this issue. We respectfully request that this letter be entered into the hearing record, and we welcome the opportunity to provide additional information to the Committee. Respectfully submitted, Shawn M. Vasell, Vice President Corporate Affairs. ______ American-Made 5G Coalition July 30, 2026 Hon. Deb Fischer, Chair, Hon. Ben Ray Lujan, Ranking Member, Subcommittee on Telecommunications and Media, Committee on Commerce, Science, and Transportation, United States Senate, Washington, DC. Re: Letter for the Record--Hearing on ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications'' (July 30, 2026) Dear Chair Fischer and Ranking Member Lujan: The American-Made 5G Coalition brings together U.S. manufacturers, wireless equipment makers, systems integrators, and rural providers. We write to raise two points. First, applications running on the Citizens Broadband Radio Service (CBRS) already benefit from artificial intelligence today on the floors of American factories. Second, AI can now make the sharing of that band better and more efficient, increasing the capacity of spectrum the country has already allocated. Neither point costs the government anything, and neither requires new spectrum. Both require that the band stay where it is. Start with what is already working. Three of every four private 5G networks in this country run on CBRS, the 3.55 to 3.7 GHz band the FCC opened to commercial use in January 2020. More than 2,500 are projected to be operating in American factories by 2032, better than 85 percent of them on CBRS.\1\ Manufacturers use them for exactly the kind of AI this hearing concerns, for example cameras that inspect parts coming off the line and models that predict which machine will fail next.\2\ That work can stay on a private network so the data never leaves the building. The payoff shows up in reductions in downtime, which costs American manufacturers about $50 billion a year, roughly $137 million a day, and which one automaker cut by 30 percent after installing a CBRS network.\3\ --------------------------------------------------------------------------- \1\ Analysys Mason, The Role of Private 5G and CBRS in the U.S. Manufacturing Sector 1, 11, 15 (Mar. 2026), https:// www.americanmade5g.com/_files/ugd/3e8afa_d81de8f298404bfa9df25 c299b818dfb.pdf. \2\ Id. at 7-8 & fig.3.2. \3\ Id. at 1, 16-17 (attributing the $50 billion estimate to a 2022 Forbes analysis). The daily figure is derived from the annual estimate. --------------------------------------------------------------------------- Now consider how the CBRS band works. A computer decides who uses it. Government radars come first, license holders second, and everyone else operates license-free on whatever remains. An automated coordination mechanism makes those assignments continuously and revises them as conditions change.\4\ That design makes CBRS unusual among American spectrum. Every assignment, every interference event, and every deployment outcome gets recorded. The band generates a detailed account of itself. --------------------------------------------------------------------------- \4\ 47 C.F.R. pt. 96 (2025) --------------------------------------------------------------------------- That record is why AI matters here more than almost anywhere else in spectrum policy. Sharing is at bottom a computing problem: who transmits where, at what strength, without stepping on anyone else. Last month, one of our member companies released a system that puts AI at the heart of making sharing in the band even more efficient. Instead of automating paperwork, the system models how radio signals actually behave in the physical world, and it learned to do so from analyzing ten years of one CBRS clearinghouse's records. Measured across hundreds of live commercial CBRS networks, the company is yielding five times more network capacity using existing spectrum and infrastructure, over 90 percent accuracy in network planning, and reducing network costs by up to 40 percent.\5\ --------------------------------------------------------------------------- \5\ Press Release, Federated Wireless, Federated Wireless Introduces Spectrum AI, the First Physical AI Platform for Shared Spectrum, Delivering Up to 5X More Capacity Using Existing Spectrum and Infrastructure (June 18, 2026), https://www.globenewswire.com/news- release/2026/06/18/3314223/0/en/federated-wireless-introduces-spectrum- ai-the-first-physical-ai-platform-for-shared-spectrum-delivering-up-to- 5x-more-capacity-using-existing-spectrum-and-infrastructure.html (reporting figures the company describes as field-validated across hundreds of live commercial deployments, and describing the platform as trained on more than a decade of spectrum assignments, interference events, and propagation observations). --------------------------------------------------------------------------- For those reasons, the CBRS band must be preserved. A decade of coordination records is what trained the models that can improve CBRS. Move the band, or change the underlying rules by increasing power levels, and that record describes spectrum no one operates in anymore. The country would be starting over on dynamic spectrum sharing innovation--a goal of this country for decades--at the precise moment it is being advanced, and it would be starting over in the one stretch of spectrum where American companies build the gear and American factories run the AI. We would welcome the chance to help you and your staff. Respectfully submitted, John Puskar, Chief Executive Officer, American-Made 5G Coalition. cc: The Honorable Ted Cruz, Chairman, Committee on Commerce, Science, and Transportation The Honorable Maria Cantwell, Ranking Member, Committee on Commerce, Science, and Transportation Members of the Subcommittee on Telecommunications and Media ______ Fiber Broadband Association Hon. Deb Fischer, Chair, Subcommittee on Telecommunications and Media, Committee on Commerce, Science, and Transportation, U.S. Senate, Washington, DC. Hon. Ted Cruz, Chair, Subcommittee on Telecommunications and Media, Committee on Commerce, Science, and Transportation, U.S. Senate, Washington, DC Hon. Ben Ray Lujan, Ranking Member, Subcommittee on Telecommunications and Media, Committee on Commerce, Science, and Transportation, U.S. Senate, Washington, DC. Hon. Maria Cantwell, Ranking Member, Subcommittee on Telecommunications and Media, Committee on Commerce, Science, and Transportation, U.S. Senate, Washington, DC. RE: Letter for the Record, July 30, 2026, Senate Commerce, Science and Communications Subcommittee on Telecommunications and Media Hearing: ``Intelligent Networks: Powering Artificial Intelligence and Transforming Communications'' Dear Chairman Fischer and Cruz and Ranking Members Lujan and Cantwell: On behalf of the Fiber Broadband Association (FBA), thank you for the opportunity to submit this statement for the congressional record. We respectfully request that the accompanying white papers, The Fourth Pillar of the AI Era and Building the Nervous System of a Thinking Economy, be included as part of the hearing record. These papers examine the critical role that fiber infrastructure will play in enabling artificial intelligence, powering intelligent networks, and sustaining America's economic competitiveness, technological leadership, and national security. Artificial intelligence is rapidly transforming every sector of the economy. To date, much of the national discussion has appropriately focused on semiconductors, advanced computing, foundation models, and energy. However, these innovations cannot achieve their full potential without the high-capacity, ultra-low-latency fiber infrastructure that connects data centers, cloud platforms, edge computing facilities, enterprises, and communities. Fiber is no longer simply the foundation of broadband--it is becoming the essential infrastructure that enables intelligence to move, scale, and create value across the economy. The accompanying studies highlight both the extraordinary opportunity and the urgency before the United States. In 2025, the fiber industry passed a record 11.8 million U.S. homes, bringing total fiber passings to more than 98 million nationwide\1\. At the same time, AI infrastructure is evolving at an unprecedented pace. Next-generation AI systems are approaching bidirectional bandwidth requirements of 3 to 4 terabytes per second per chip, driving exponential demand for optical networking, data center interconnection, middle-mile infrastructure, and last-mile fiber connectivity.\2\ Meeting these demands will require a national policy framework that recognizes fiber as strategic infrastructure and accelerates its deployment. --------------------------------------------------------------------------- \1\ Fiber Broadband Association, ``Fiber Broadband Association Reports Record Fiber-To-The-Home Deployment in 2025,'' January 2026. Reports 11.8 million U.S. homes passed in 2025 and 98.4 million total U.S. fiber passings including multiple passings. \2\ Fiber Broadband Association, The Fourth Pillar of the AI Era, April 2026, p. 2 --------------------------------------------------------------------------- As Congress considers policies to strengthen America's leadership in artificial intelligence and next-generation communications, the Fiber Broadband Association respectfully recommends two priorities: First, recognize fiber infrastructure as a foundational pillar of the Nation's AI strategy. Alongside compute, models, energy, and data, fiber provides the high-capacity, low-latency connectivity required for AI systems to operate reliably, securely, and at scale. Without robust fiber infrastructure, the full economic and societal benefits of artificial intelligence cannot be realized. Second, modernize and streamline permitting processes to accelerate deployment of the fiber infrastructure required to support AI and intelligent networks and emerging technology innovation. Predictable, transparent, and timely permitting policies will reduce unnecessary delays, encourage continued private-sector investment, and ensure the United States can deploy the infrastructure necessary to remain globally competitive in the AI era. The United States has a unique opportunity to lead the world not only in developing artificial intelligence, but also in building the infrastructure that enables it. The decisions made today regarding fiber investment and deployment will shape America's innovation capacity, economic competitiveness, and technological leadership for decades to come. The Fiber Broadband Association appreciates the Committee's leadership in examining the infrastructure requirements of the AI era and respectfully submits these white papers for inclusion in the congressional record. We stand ready to work with Congress and the Administration to advance policies that ensure the United States continues to lead in both artificial intelligence and the fiber infrastructure upon which it depends. Sincerely, Gary Bolton, President and CEO, Fiber Broadband Association. Senator Fischer. Senator Lujan will be joining us shortly. But we will begin the hearing, and I would like to introduce our witnesses for today. Our first witness is Jonathan Spalter, President and Chief Executive Officer of USTelecom--The Broadband Association, a trade association representing the broadband providers and communication companies that build, operate, and secure networks essential for America's AI economy. I would like to give a warm welcome to our second witness, Commissioner Dan Watermeier of the Nebraska Public Service Commission, which oversees the telecommunication carriers and related services that support AI-powered applications throughout the state of Nebraska. Our third witness is Bob Everson, Chief Architect of Provider Mobility at Cisco, where he leads the design and development of mobile network platforms and end-to-end architectures that enable AI-driven communications networks. And our final witness is Asad Ramzanali--is that correct, Ramzanali--Director of Artificial Intelligence and Technology Policy at the Vanderbilt Policy Accelerator at Vanderbilt University. Welcome to you all. And we will now have the opening comments from our panel. Mr. Spalter, I understand it is your birthday. Happy birthday. Mr. Spalter. Well, thank you very much, Senator. Senator Fischer. And you are going to be recognized for five minutes. Mr. Spalter. Thank you. And I couldn't imagine spending my birthday anywhere else but here. But thank you. Senator Fischer. I gather so. STATEMENT OF JONATHAN SPALTER, PRESIDENT AND CHIEF EXECUTIVE OFFICER, USTELECOM--THE BROADBAND ASSOCIATION Senator Fischer. But thank you, Chair Fischer and Senator Klobuchar, as well as to the other Members of the Subcommittee. When I was thinking about how to prepare my remarks today, I took some inspiration, actually, from a trip I took a couple of weeks back to the rafters of the Capitol Dome to take a look at Constantino Brumidi's amazing Fresco. I looked up, and I saw, much to my disbelief, given my day job, that there was the Goddess Venus actually laying the very first American submarine telegraph cable. The scene was actually painted in 1865, the very moment that America was laying that very first transoceanic telegraph cable, the Civil War was ending. And it was then that a very visionary Congress believed that the future of our country ran through communications infrastructure. Well, they were right then, and you on this committee, are so right now to expand that dialogue in the current AI moment that we are in. Because with AI, we really are at the cusp of the next fundamental evolution of the internet. For 30 years, the Internet we have come to know was built kind of like a freeway, imagine with eight lanes going into town but only two lanes coming back. And that asymmetric design was literally baked in as the assumption that--on the assumption that we, human beings, and our devices would always receive more data than we send. And as we all go home to binge stream Love Island while our kids are playing Roblox, that is very still much the case. Except that as we move forward into this AI phase of the internet, that assumption is being completely shattered. Now the machines and sensors all around us require massive capacity to carry data both upstream and down, which fiber broadband uniquely is suited to do. Other technologies we know are working to expand their upstream capacity, but fiber doesn't have to, it never had to. It was built for this era with near limitless capacity and low latency in both directions. And this is why satellite, and wireless, and Wi-Fi, and other technologies that might become available in the connectivity landscape will always hand off their traffic to fiber as quickly as they can to take advantage of its massive capacity and low latency. And it is for this reason that fiber broadband really is the circulatory system of our AI-enabled innovation economy. Thus, for us, the biggest barrier to getting AI-ready connectivity to every community isn't technology, it isn't investment, nor is it our will. We are ready. It is government. Specifically permitting, waiting months, even years for permission to connect more communities to broadband. So our asks today are very straightforward. Pass the CLOSE THE GAP Act, the RAIL Act, and the Accelerating Broadband Permits Act, so we can get our shovels and our fiber network in the ground, and connect more of your communities to make sure that they, too, are going to be part of our AI future. And let me emphasize this point. Without permitting reform to speed up fiber network deployment, not just AI, but our whole innovation economy is at risk of being put on ice. And nowhere is this more true than in rural America, as you rightly pointed out, Chair. The communities that need us most are those where your actions today can make the biggest difference. In 1865, Congress looked at a single telegraph cable on the ocean floor and they saw the American future. We were optimists then, and we must be optimists now with a healthy dose of pragmatism built in as well. So the question before us, I believe as a society, is not do we stop or continue, it is the question that has been perpetually asked by innovation, one that we have successfully answered: How do we proceed responsibly in ways that will most fully benefit everyone? And that is the task before us. America's broadband providers, to all persons, look forward to working with you. And on behalf of all of USTelecom member companies, I really thank you for the opportunity to be here today on my birthday. I look forward to your questions. [The prepared statement of Mr. Spalter follows:] Prepared Statement of Jonathan Spalter, President and CEO, USTelecom Chairman Fischer, Ranking Member Lujan, and members of the subcommittee: thank you for the opportunity to appear before you today. I am Jonathan Spalter, President and CEO of USTelecom--The Broadband Association. Our members are the companies that build, own, operate, and defend the broadband networks that reach into nearly every community in this country, from the largest cities to the Nebraska Sandhills and the mesas of northern New Mexico. I want to begin by suggesting a frame for how we think about AI. We have a habit, in the early years of any transformative technology, of treating it as a category of its own. We once spoke of ``e-commerce'' and ``e-health'' and ``e-learning''--until they quietly became commerce, health care, and education. Artificial intelligence is on that same path, albeit fast-tracked. AI will not simply be a sector of our economy. It is going to become a layer, embedded in nearly everything: in agriculture, in logistics, in medicine, in manufacturing, in national defense, in the way a small business keeps its books and finds its customers. Artificial intelligence is not the biggest thing since the internet; it is the next, most powerful evolution of the Internet itself. Every serious account of the AI buildout now names three core ingredients: compute, energy, and connectivity. The first two dominate the headlines. The third--the networks that move the data--is the one that determines whether any of this progress reaches businesses, consumers, and government offices across America. If I could leave this subcommittee with one primary contribution today, it is this: America's AI future and the future of our entire innovation economy runs through fiber broadband networks. Not alongside them. Not around them. Through them. Fiber is the foundation on which all next-generation connectivity depends. This includes satellite and wireless, which rely on fiber to move the massive amounts of AI-fueled data that help deliver the real-world benefits of the innovation emerging all around us--to communities throughout our Nation. The primary constraint on America's AI infrastructure is not technological. It is governmental. Permitting is the single largest roadblock to delivering AI-ready connectivity in every community, and it is squarely within Congress's power to address this year. Broadband providers are ready to build connections across this country, and we have the record to prove it. We have invested $2.2 trillion over the past 30 years, since the earliest days of the commercial internet. We directly employ roughly 570,000 Americans-- engineers, data scientists, and cybersecurity specialists working alongside the field technicians who keep these networks running,\1\ and we have decades of experience working with artificial intelligence to strengthen the reliability and efficiency of our networks. --------------------------------------------------------------------------- \1\ U.S. Bureau of Labor Statistics, Current Employment Statistics: Wired and Wireless Telecommunications Carriers (Except Satellite), NAICS 517111 (2024). This figure reflects broadband providers' direct employees; it excludes the construction and contractor trades that build and maintain networks, as well as satellite carriers. --------------------------------------------------------------------------- Our task is to advance constructive solutions that allow us to work together, policymakers and the private sector, to connect every community. Whether a farmer in Custer County, a machine shop in Grand Island, or a nurse practitioner in the only clinic for sixty miles ever sees the benefit of any of this progress depends entirely on the power of the network at their door. So, I hope a central question for us is not who builds the computing, but whether every American will be connected well enough to use it. I. Fiber Is the Foundation of the AI Economy Let me start with the physical reality that underlies everything else: AI requires the continuous movement of massive amounts of data, and how much data can move is a function of the medium it moves through. No medium moves more than fiber. When a community competes to land an advanced manufacturing plant, when a regional health system builds out remote diagnostics, those demands rely on access to fiber. Fiber is the scalable backbone that every Internet technology depends upon. It is, in the truest sense, the circulatory system of the American innovation economy. Without question, every technology has an important role in the future we are shaping together. Fixed and mobile wireless and satellite serve real and important purposes--reaching the hardest-to-reach geographic areas, backfilling aging copper, connecting ships and aircraft, restoring service in the aftermath of a disaster. But fiber is the infrastructure foundation for our AI future. A fiber optic strand is thinner than a human hair. Buried in the ground and built to last for decades, it has three distinct advantages. Capacity that scales. The fiber itself is not the limiting factor: capacity typically can be expanded simply by upgrading the electronics at either end, not digging back into the ground. So, the same fiber laid today can carry many times the traffic of today's internet, and stands most capable among all technologies to keep pace with what AI will demand tomorrow. Backhaul. Every other Internet technology relies on fiber. 5G and 6G cell sites connect to fiber. Satellites land their traffic at ground stations that connect to fiber. Latency. The closer we get the fiber to the home or to the business, the faster and more capable that connection will be. On a video call, a few dozen milliseconds is invisible. For a robotic arm working beside a person on an assembly line, for a vehicle deciding whether to brake, for a surgical instrument guided from another hospital, timing is everything. Putting fiber in the ground requires a skilled workforce-- surveyors, trenching crews, splicers, and technicians who live in the communities they connect. These are jobs in your states: trenching crews in Nebraska, splicers in New Mexico, technicians in Minnesota, Illinois, Michigan, Wisconsin, Montana, and Ohio. Demand already outstrips supply. Industry analyses already project a shortfall approaching 180,000 skilled workers through the early 2030s.\2\ Not every technology now competing for funding creates work of this kind, in our communities. --------------------------------------------------------------------------- \2\ Patience Haggin, ``High-Speed Internet Boom Hits Low-Tech Snag: A Labor Shortage,'' Wall Street Journal, February 1, 2026, https:// www.wsj.com/business/telecom/high-speed-internet-boom-hits-low-tech- snag-a-labor-shortage9c92b514 --------------------------------------------------------------------------- The same is true at the community level. As our Nation works to bring the production of chips, routers, microprocessors and optical transceivers back to the United States, we know advanced manufacturing does not locate where the network does not reach. Reindustrialization and broadband deployment are the same project, and the rural communities competing for those facilities know it. II. Making Sure Every American Can Participate AI adoption is not just a ``big tech'' story. It is a Main Street story. The JPMorgan Chase Institute finds that 17.7 percent of small businesses had purchased an AI service in 2025, more than triple the share two years earlier. This is almost certainly an undercount, since it captures only direct payments, not the free or embedded tools many use today.\3\ The U.S. Chamber puts generative AI use among small businesses at 58 percent, up from 23 percent in 2023.\4\ The work is routine and practical--whether it's drafting the estimate or the marketing e-mail, answering the customer who calls after hours, keeping the books, or flagging equipment that may be in need of maintenance.\5\ Nearly two-thirds of foundations and nonprofits report using AI.\6\ And 80 percent of small businesses say AI enhances their workforce, with 82 percent of those using it having grown their headcount over the past year.\7\ --------------------------------------------------------------------------- \3\ JPMorganChase Institute, Understanding AI Use by Small Businesses (2026), https://www .jpmorganchase.com/institute/all-topics/business-growth-and- entrepreneurship/understanding-ai-use-bysmall-businesses. \4\ U.S. Chamber of Commerce, Empowering Small Business: The Impact of Technology on U.S. Small Business (2025), https://www.uschamber.com/ technology/empowering-small-business-the-impact-of-technology-on-u-s- small-business. \5\ U.S. Chamber of Commerce Foundation, AI in Action: Early Lessons from Small Businesses on the Front Lines of Adoption (2026), https://www.hiringourheroes.org/resources/ai-in-action-whitepaper/. \6\ Center for Effective Philanthropy, AI With Purpose: How Foundations and Nonprofits Are Thinking About and Using Artificial Intelligence (2025), https://cep.org/report-backpacks/ai-with-purpose- how-foundations-and-nonprofits-arethinking-about-and-using-artificial- intelligence/ \7\ Goldman Sachs small business survey, as reported in Fox Business (August 8, 2025), https://www.foxbusiness.com/economy/small- business-ai-adoption-jumps-68-owners-plan-significant-workforcegrowth- 2025; U.S. Chamber of Commerce, Empowering Small Business: The Impact of Technology on U.S. Small Business, supra note 4. --------------------------------------------------------------------------- The opportunities created by these advances extend beyond the private sector. In New Mexico, a public university now uses AI across its admissions office--answering prospective students around the clock, in multiple languages, and freeing staff for person-to-person counseling work. What unites the companies USTelecom represents is a commitment to connect every community to the best technology we can get to them. That is why we are such large investors ourselves in the Nation's broadband infrastructure, and why we have worked with this body on programs like BEAD that can bridge the cost gap where private capital alone cannot reach remote areas. The communities with the most to gain from AI are very often the ones where the economics are toughest. Which is the heart of the matter: the places where private capital struggles hardest are the places where your leverage is greatest. To those communities, a permit granted in 90 days instead of three years is worth more than any promise the rest of us can make them. III. Broadband Companies Have Been Using AI in Our Networks for Decades The Subcommittee asked how the communications sector is using AI tools to run networks more efficiently, anticipate outages, and defend against cyberattacks. The industry's experience here is not new. As far back as the 1980s, telephone companies deployed rule-based ``expert systems'' for fault management--flagging when something on the network failed, isolating the cause, and clearing it. Machine learning followed, for call routing, predictive monitoring, and fraud detection. Today our networks can increasingly diagnose and fix themselves, and some carriers now train AI models on their own networks, so the system learns the quirks of that particular network the way a seasoned engineer learns the one she has run for years--except across millions of readings at once, continuously, and without ever looking away. These are not hypotheticals: One major broadband provider reports making roughly 700,000 changes to its network every day with AI, using a foundation model built for its own network. Another now runs some 200 edge computing centers that automatically resolve more than three- quarters of network events before a customer is affected, locating faults with better than 99 percent precision.\8\ --------------------------------------------------------------------------- \8\ Phil Harvey, ``The AI-Enabled Network Is Here. The Pitch Is Stuck in Traffic,'' Light Reading (May 29, 2026), https:// www.lightreading.com/ai-machine-learning/the-ai-enabled-network-is- here-the-pitch-is-stuck-in-traffic --------------------------------------------------------------------------- Network efficiency. Demand fluctuates enormously across a single day--the morning surge as a region logs on, the evening crest of video streaming, a stadium that fills on a Saturday. AI systems proactively watch the network in real time and steer traffic around congestion, dynamically allocating capacity where it is needed most. The same systems increasingly anticipate outages before they happen. This is the shift from reactive repair to predictive resilience. AI models can identify the small anomalies that precede a failure--a degraded optical signal, an irregular power draw, a component drifting out of alignment--so a provider can dispatch a technician or reroute traffic before customers are affected. The outage that never happens is invisible to the public. It is one of the most valuable things AI does in our networks. The same logic scales up to disasters: Leading broadband providers use AI to model approaching hurricanes to predict which assets are most likely to be hit, so crews, generators, and portable capacity can be positioned before landfall. Cybersecurity. Our broadband sector is keenly aware that America's adversaries are using this same technology to attack our infrastructure at greater speed and scale. That is why USTelecom is a contributor to the National Institute of Standards and Technology's work on AI as part of its Cybersecurity Framework. This is a dynamic, industry-informed approach where policy leaders and broadband experts roll up their sleeves together on the kinds of solutions these complex challenges call for.\9\ --------------------------------------------------------------------------- \9\ See National Institute of Standards and Technology, Cybersecurity Framework Profile for Artificial Intelligence (Cyber AI Profile): NIST Community Profile, NIST IR 8596, initial preliminary draft (December 2025), https://csrc.nist.gov/pubs/ir/8596/iprd. --------------------------------------------------------------------------- Network engineers now work alongside AI agents that read the flood of alarms coming off the network, pinpoint the real problem, open a ticket, propose a fix, and write the incident summary. These capabilities run in parts of our members' networks, not yet across all of them, and extending them takes expanding our fiber capacity, which again brings us back to permits. IV. AI Is Changing the Shape of Network Traffic As we look at the growth in network traffic attributable to AI, the volume is striking. Consumer network traffic is forecast to run roughly 6.6 times higher by 2035, with AI the dominant driver of growth. Agentic AI for enterprise is expected to bend the curve even more steeply, powering a 9fold expansion in related traffic by 2035.\10\ --------------------------------------------------------------------------- \10\ Cisco, AI Impact on Wide Area Networks (2026), https:// www.cisco.com/c/dam/en/us/solutions/collateral/artificialintelligence/ mass-scale-infrastructure/ai-network-traffic-report.pdf. --------------------------------------------------------------------------- Of course, ten years from now might as well be a century, given how quickly the landscape is evolving. The pace is already visible: this June, for the first time in the history of the internet, automated systems generated more search requests than human beings--57.4 percent from bots and AI agents, against 42.6 percent from people.\11\ Experts can differ in good faith on the precise forecasts, but the more consequential change gets less attention: AI is not simply increasing the volume of network traffic. It is changing the shape of it. And the shape must determine what we build, and where. --------------------------------------------------------------------------- \11\ Cloudflare data reported, NBC News, Bot Web Traffic Has Overtaken Human Web Traffic, Data Shows (June 3, 2026), https:// www.nbcnews.com/tech/tech-news/bot-web-traffic-overtaken-human-web- traffic-data-showsrcna348522. Traffic is reversing direction. Two terms are important to understand: upstream is everything traveling toward the computing--the prompt, the photograph, the sensor reading; downstream is everything coming back--the answer, the diagnosis, the instruction. For 30 years the Internet has been engineered overwhelmingly for downstream. Picture a freeway built with eight lanes running into town and two running out. That is what was needed to meet the network demands of video streaming. But AI inverts the flow, pushing enormous volumes of data back upstream: rich prompts, video, sensor and lidar feeds that allow machines to register their surroundings. As an example, a single connected vehicle generates roughly 20 gigabytes of data per day.\12\ This is roughly 30 times what the average mobile customer consumes, and it flows in the direction most network technologies--copper, coaxial cable, wireless, satellite--were least designed to carry. Fiber is the exception: the same, high-capacity lanes in both directions. --------------------------------------------------------------------------- \12\ U.S. Department of Transportation, Federal Highway Administration, Maricopa Association of Governments Finds Value in Connected Car Data, Crowdsourcing for Operations Case Study, FHWA-HOP- 22-047 (2022), https://www.fhwa.dot.gov/innovation/everydaycounts/ edc_6/docs /crowdsourcing_mag_case_study.pdf. Traffic is becoming ``bursty'' and machine-driven. An AI agent does not browse the way a person does. A person loads a page, reads, thinks, clicks. An agent fires off dozens of requests in a second, goes quiet, then does it again--short, intense spikes rather than a steady stream, and a network sized for the steady stream is generally not sized for the spike. Cisco finds a 450 percent increase in traffic when a task is executed by an AI agent rather than a human.\13\ An AI chatbot says, ``Here is an answer.'' An AI agent says, ``Here is the plan, and I will carry it out step by step.'' Every one of those steps is traffic. --------------------------------------------------------------------------- \13\ Cisco, AI Impact on Wide Area Networks (2026), supra note 10. Traffic is becoming intolerant of delay--and geographically distributed. This is the change that most directly implicates Federal policy. There are machines acting in the physical world, on a deadline. A traffic camera using AI to prevent a collision cannot wait for a round trip to a server several states away. An autonomous system on a factory floor has no margin for a lagging signal. These applications require computing and connectivity close to where the work is happening. And that means more fiber, deeper into more places, to directly support communities and to support the wireless, --------------------------------------------------------------------------- satellite and other connectivity options that ride on it. The defining infrastructure challenge of the AI era is not simply how much we build. It is where, with what, and how fast. Latency is a function of distance, and distance is a function of geography. You close it by building closer to the user--which means opening a trench, crossing a right-of-way, filing a permit. Thousands of times, in thousands of jurisdictions. Simply put, the physics of AI turns a computing challenge into a permitting challenge. And, that's a problem we can solve today. V. The Broadband Community Is Doing Its Part Consider the scale of what private capital has already built: $2.2 trillion invested by broadband companies since 1996, and $89.6 billion in 2024 alone\14\ with 2025 expected to be higher still--sustained through recessions, higher interest rates, and a cautious capital environment. To put that in context, Congress committed $42.45 billion for BEAD, the largest Federal broadband infrastructure program in U.S. history. It is designed to fill a critical gap: connecting the communities that cannot otherwise be reached, where the cost of building broadband networks exceeds any realistic private return. Roughly half of BEAD funds, about $21 billion, are going to deployment. This is about one percent of the private capital the industry has invested since 1996. America's digital infrastructure is overwhelmingly privately financed, and it was built because both parties, across many administrations, sustained a pro-investment posture: stable rules, predictable policy, and a settled preference for building. --------------------------------------------------------------------------- \14\ USTelecom, 2024 Broadband Capex Report (October 21, 2025), https://ustelecom.org/research/2024-broadbandcapex-report/. --------------------------------------------------------------------------- VI. The Limiting Factor Is Government, Not Technology Broadband providers are ready to build. What stands in the way is a permitting system designed for a different era, and a problem that is often misdiagnosed. Our difficulty is rarely an outright ``no'' from a permitting authority. Our primary difficulty is the absence of an answer. It is delay. Running out the clock--through a construction season, through a grant deadline, through a community's patience--until the project is more expensive or abandoned. Among our members' experiences: Illinois. A provider may hold only one active permit at a time; each must be closed out before the next is released. Maryland. A city stalled a provider's effort to install a small fiber hut for nearly a year by requiring an extensive zoning review typically reserved for large commercial developments. Minnesota. A city demanded a $63,000 permit fee plus nearly $29,000 in per-foot charges for a single block of fiber. When it refused a compromise, the provider had to walk away. Hawaii. The legislature enacted a 60-day broadband shot clock nearly a decade ago. Not a single permit has ever been processed in that timeframe. Ohio. One town simply ignored a provider's application. The deployment was abandoned. Montana. A member received a Federal grant in June 2023 and, two years later, had not turned a single shovel of dirt-- waiting on permits. The Mountain West. One provider waited three years for a permit in an area with a six-week annual construction window, narrowed by weather, terrain, and protected-species breeding seasons. By the time the permit arrived, the season had closed. Federal lands. Permitting can take up to four years. One member was made to conduct a fresh environmental review of a corridor that already held a telephone line, power and natural gas lines, and a county road--land previously disturbed and previously analyzed. Big Bend National Park. A member waited two years to bring fiber to homes and businesses inside the park, held up in DC by a required review of forest resources. The problem? Big Bend sits in the Chihuahuan Desert. There is no forest. Railroads. It is nearly impossible for a fiber deployment not to be stopped in its tracks by a railroad demanding fees that bear no relation to cost to bore or cross under a railroad. Congress, and this committee, should pass the RAIL Act to put in place common-sense rules that preserve safety for railroads while limiting the delays that serve no safety purpose. Permitting officials are doing important work, most of them under resourced. And when the rules are clear and followed, communities win. This leads to my most important point: We are not speculating about whether permitting reform works. We know that it does. In 2018, as part of the same legislative push that cleared the runway for 5G, Congress required Federal agencies to grant or deny communications permits on Federal land within 270 days.\15\ Where agencies kept reliable records, the Government Accountability Office found average processing time fell by 57 percent between 2018 and 2022. GAO's other finding is just as important: the Bureau of Land Management could not reliably determine its own processing time for 42 percent of applications.\16\ Congress set the deadline. Some agencies still cannot confirm they are meeting it. Both halves need fixing--and the CLOSE THE GAP Act would do just that. --------------------------------------------------------------------------- \15\ Consolidated Appropriations Act, 2018, Pub. L. No. 115-141, div. P, tit. VI (MOBILE NOW Act), Sec. 606 (establishing timelines for Federal agency action on communications facility siting applications on Federal property). \16\ U.S. Government Accountability Office, Broadband: Agencies Should Improve Data Reliability and Coordination for Permitting on Federal Lands, GAO-24-106157 (2024), https://www.gao.gov/products/gao- 24-106157. Also in 2018, the FCC adopted shot clocks and cost-based fee caps for small wireless facilities to more densely cover a neighborhood, a corridor, a stadium. Small-cell deployment then grew by 110 percent.\17\ This resulting densification is what turned 5G into the powerful network Americans now carry in their pockets. Importantly, the FCC today is working to advance a similar approach to streamline permitting for wired networks. --------------------------------------------------------------------------- \17\ Accelerating Wireless Broadband Deployment by Removing Barriers to Infrastructure Investment, Declaratory Ruling and Third Report and Order, WT Docket No. 17-79, 33 FCC Rcd 9088 (2018); CTIA, 2025 Annual Survey Highlights at 7 (2025), https://api.ctia.org/wp- content/uploads/2025/09/2025-Annual-Survey-Highlights.pdf. More recently, state Broadband Ready Community certifications for towns and counties that commit to a single point of contact, a firm review deadline, and reasonable, cost- based fees have demonstrated something important: faster reviews and reasonable fees do not cost communities anything. They signal to providers that a community is open for business. More than 280 towns, cities, and counties across Georgia, Indiana, Tennessee, and Wisconsin have adopted these --------------------------------------------------------------------------- certifications. The Subcommittee will notice that some of these reforms date to 2018--and that is no accident. They were built for wireless: for the small-cell densification that made 5G possible. Nothing on the same scale has ever been done for wireline. The playbook exists. It worked once. It has simply never been applied to the infrastructure that matters most today. Every delay is a constituent community that waits. And every delayed broadband build is a delay in strengthening America's AI infrastructure. Computation without connectivity reaches only those who can afford to build their own onramps. Our job, and yours, is to make sure it reaches everyone else, too. VII. What We Ask of Congress 1. Pass the CLOSE THE GAP Act (S. 4561). Introduced by Senators Barrasso and Lummis, this bill goes directly at the federal- lands bottleneck: it streamlines agency permitting regulations for broadband and exempts previously permitted and previously analyzed land from duplicative review. The communities waiting on the far side of a four-year permit are found in New Mexico as surely as in Wyoming, Montana, and Nebraska. It is targeted, it is sensible, and it is ready.\18\ --------------------------------------------------------------------------- \18\ CLOSE THE GAP Act, S. 4561, 119th Cong. (2026) (introduced by Sen. Barrasso, cosponsored by Sen. Lummis), https://www.congress.gov/ bill/119th-congress/senate-bill/4561. 2. Pass the Broadband and Telecommunications RAIL Act (S. 3268). Introduced by Senators Blackburn and Lujan, this bill goes directly at the railroad-crossing bottleneck: it sets enforceable timelines for railroads to respond to and schedule crossing work, reins in the arbitrary fees that can run into the tens of thousands of dollars for a single crossing, narrows the grounds for denial to genuine questions of safety and railroad operations, and gives the FCC a role in resolving disputes. It does all of this while preserving the safety standards that rail carriers rightly require, since providers must still submit engineering plans and coordinate closely with the railroads. The bottleneck here is not safety. It is a permit that can sit unanswered for the better part of two years while the town on the far side of the tracks waits for service.\19\ --------------------------------------------------------------------------- \19\ Broadband and Telecommunications RAIL Act, S.3268, 119th Cong. (2026) (introduced by Sen. Blackburn, cosponsored by Sen. Lujan), https://www.congress.gov/bill/119th-congress/senate-bill/3268. 3. Pass the Accelerating Broadband Permits Act (S. 4448). Introduced by Senators Thune, Lujan, and Barrasso, this bill goes directly at the accountability bottleneck. It cuts duplicative red tape in agency review and brings transparency to where an application sits in the process. Congress set that deadline in 2018. GAO reported in 2024 that agencies were still missing it. A deadline without a mechanism is only a suggestion, and the households waiting on an unanswered file are in South Dakota and New Mexico as surely as in Wyoming and Nebraska. 4. Reauthorize CISA 2015. Finally, to ensure the safety and security of the networks we build, it is imperative that Congress reauthorize CISA 2015 so that we can continue to safely, quickly, and effectively share information across our industry and with our government partners. Conclusion Chair Fischer, Ranking Member Lujan: at the end of all of this are families in small towns across the country who want to know what any of it means for them--for their quality of life, for their kids, for the cost of running their business. The honest answer is that everything we have described arrives in their lives only if there is a network at their door capable of delivering it. America's broadband providers have built these networks across a generation, at a scale few industries can match, and we are ready to keep building. We are not asking Congress to build them for us. We are asking you to clear the path--and, where the economics will not reach, to keep the Federal partnership strong enough to finish the job. The gateway is government. The levers that remain are the ones you hold. Pull them, and the powerful modern networks through which America's future transits will be built faster than anyone expects--and every community will be on the other end of the opportunities they make possible. Thank you. I look forward to your questions. Senator Fischer. Thank you, Mr. Spalter. We have been joined by my dear friend and colleague, Ranking Member, Senator Lujan. Did you have comments that you would like to make at this time? STATEMENT OF HON. BEN RAY LUJAN, U.S. SENATOR FROM NEW MEXICO Senator Lujan. Madam Chair, thank you so much for putting this hearing together, and to the witnesses for being here today. Look forward to the testimony. [The prepared statement of Senator Lujan follows:] Prepared Statement of Hon. Ben Ray Lujan, U.S. Senator from New Mexico Thank you to Chairman Fischer, Chairman Cruz, and Ranking Member Cantwell for calling this hearing. Thank you to our witnesses for being here today to discuss the role of artificial intelligence in the buildout and management of our communications networks. AI is transforming the way our networks are built, managed, and secured--and it is doing so in a way that is reshaping the digital economy, our global competitiveness, and the development of advanced and secure networks. As we deploy broadband across the country, we need to think strategically about how today's investments can support tomorrow's technologies: building fiber networks and other infrastructure that can serve communities now while creating the capacity needed to support emerging AI and quantum applications. This approach can help ensure that rural and underserved communities are not left behind in the innovation economy, but instead gain connectivity, workforce opportunities, and access to advanced research and industry that can help them grow. The question today is not whether AI will transform communications, but instead whether that transformation reaches all Americans and not just the wealthiest. I want to state a simple fact: AI runs on broadband infrastructure. This means every AI-enabled service or application depends on physical networks to reach people. I believe that this infrastructure must include rural and Tribal communities. In New Mexico, that means laying the groundwork for broadband that supports statewide quantum computing and artificial intelligence networks. If we want America to lead the world in AI, our networks must be built on a solid foundation of consistent policies across the Federal government. That's why ensuring that the states receive all of the $42.5 billion of Broadband Equity, Access, and Deployment, or ``BEAD,'' funding matters. BEAD funding was appropriated by Congress on a bipartisan basis, so that every state and territory could finally close the digital divide. But the Trump Administration chose a different path. It spent its first year in office rewriting the rules and eliminated the program's preference for fiber--the most reliable, most future-proof technology, and the one AI depends on. States were forced to redo work, and then slow-walked through the approval process. New Mexico was approved for $382 million in BEAD money, while the Trump Administration holds the remaining $293 million hostage. But it is not just New Mexico that is impacted by these bureaucratic delays. The Administration is currently holding on to over $21 billion in non-deployment BEAD funding. That's nearly half of the money that Congress appropriated to BEAD for broadband adoption, digital literacy, and workforce programs. States have been waiting for the non-deployment guidance for months. In December, Senator Fischer and I co-led a letter to NTIA Administrator Roth and urged NTIA to preserve the states' ability to use non-deployment BEAD funds consistent with the bipartisan infrastructure law. In March, the response was that NTIA needed additional time. Well, how much more time? Because we're still waiting. This Administration is trying to take all the credit for championing American leadership in AI, while it allows the Commerce Department to sit on billions of dollars of Congressionally appropriated broadband funding, and degrades the technology standards that AI-ready networks need. The Administration can't drag its feet on broadband infrastructure and claim leadership in AI. And this Administration's bad broadband record is not limited to BEAD delays. You cannot claim to champion AI leadership while unilaterally terminating digital literacy and digital equity funding that Congress lawfully appropriated--funding that was cut off through a social media post. That decision was legally reckless, and a Federal court recently issued an order expecting the Administration to reinstate the Competitive Grant Program--something the Administration has previously said they would do. The infrastructure needed for AI is only part of what we are exploring in today's hearing. AI has the potential to improve network resiliency, change how we defend against cyber attacks, and optimize traffic across the networks. While there are many tangible benefits to AI, I am also concerned that this technology can be used for nefarious purposes, which underscores the need for us to be diligent in our oversight and understanding of how AI is being deployed within our networks. The recent Hugging Face incident is illustrative of this concern. An AI model being evaluated in what was supposed to be a controlled testing environment, exploited an unknown vulnerability. As these models become more capable, we need to ask serious questions about how these AI systems are being tested and deployed within our infrastructure. We are fortunate to have an expert panel with us today, and I look forward to a productive conversation. Thank you all for being here. Senator Fischer. Thank you, Ranking Member. Commissioner Watermeier, you are now recognized for five minutes for your opening comments. STATEMENT OF DAN WATERMEIER, COMMISSIONER, NEBRASKA PUBLIC SERVICE COMMISSION Mr. Watermeier. Good morning, Chair Fischer, and Members of the Committee. My name is Dan Watermeier. I represent Nebraska's First District on the Public Service Commission. I appreciate the invitation to speak to you today about the future of AI technology and opportunities, and also the challenges it offers to Nebraska, and on a national scale. The Nebraska Public Service Commission is dedicated to the advancement of broadband funding. Since the early 2000s, we have administered Nebraska Universal Service Fund, which now supports the development and maintenance of broadband capable networks at 100 by 100 speeds. We have also administered the Nebraska Bridge Program and the federally funded Capital Funding--Projects Fund, which over the course of the last few grant cycles, the broadband deployment projects have fueled a competition to build over 20,000 unserved and underserved across the state. Over the past five years, we have been fortunate to receive Federal and state funds. I am excited to see how this enthusiasm will grow as we develop new AI tools. AI could become a tremendous support for our agriculture producers and already has. Precision agriculture tools are fantastic ways to conserve water, monitor livestock health, and target applications of fertilizers and pesticides, saving money and protecting the environment at the same time. These AI-powered tools are the future of agriculture, but we need reliable, universally stable and available broadband in order to use them. As adoption of and reliance on AI technology increases, it will become even more important to ensure that all Americans have access to reliable high-speed broadband networks. We need to ensure that rural Americans are not left behind on this. Technology moves forward. For the average citizen, AI technology is only useful if they have access to broadband connectivity. We must continue funding rural buildout or this technology simply won't be available for all who need it. I also want to mention the importance of fiber buildout, as my partner here had mentioned. While we recognize that many technologies may be able to deliver broadband to our citizens, our Commission has focused on fiber development for a few reasons. Specifically, first, scalability. Not every technology can easily meet the FCC's current 100/20 technology threshold. Meanwhile, fiber optic networks can greatly exceed these speed thresholds and are capable of meeting our needs into the foreseeable future. I want to point out that data centers are not using wireless or satellite technology to move data, they are using fiber. Second, supporting development. Fiber is the backbone for basically all the other technologies. We need fiber-based networks to support wireless, fixed wireless, and the cable networks. Third, reliability. Buried fiber is much less susceptible to weather hazards and natural disasters. Fires, tornadoes, and ice storms in recent years in Nebraska have shown the need for resiliency that fiber can provide. Developing broadband infrastructure will not be enough if these networks are not maintained. We have heard from industry providers that although the investment in fiber itself may last for 40 years, the security required in fiber must be upgraded much more frequently to accommodate changing technology. This to be expected as demands on the network for download and upload speeds are greatly--and latency demands continue to grow. The tasks that AI is capable of today will be eclipsed by what it can do 5 years from now, so long as the network is upgraded to keep it up. For that reason, I want to emphasize that new buildout must be as future-proof as possible in the face of developing technology. Chair Fischer and Members of this Committee, I know that you recognize the importance of maintaining broadband networks. Hopefully, it is clear that AI technology is dependent on those networks. I also think it is important to recognize that many of these networks are supported through Federal and state universal service funds. And the Nebraskans that I represent, are grateful for your work and the work of your colleagues on the USF Working Group. I look forward to hearing the group's recommendations for the future of USF. I know there is a discussion on how to Federal USF is funded. I want to just point out that AI creates and will create significant demand on the network, a network that is supported today through surcharges on voice only. Thank you for your invitation to speak. And I will gladly answer questions as you end, if I can. Thank you. [The prepared statement of Mr. Watermeier follows:] Prepared Statement of Dan Watermeier, Nebraska Public Service Commission, 1st District Good morning, Chair Fischer and members of the Committee. My name is Dan Watermeier, and I represent Nebraska's First District on the Nebraska Public Service Commission. I appreciate the invitation to speak to you today about the future of AI technology and the opportunities--and the challenges--it offers in Nebraska and on the national scale. The Nebraska Public Service Commission is dedicated to the advancement of broadband funding. Since the early 2000s, we have administered the Nebraska Universal Service Fund, which now supports the development and maintenance of broadband-capable networks at speeds of at least 100/100 Mbps. We have also administered the Nebraska Broadband Bridge Program and the federally funded Capital Projects Fund, which over the course of just a few grant cycles funded broadband deployment projects and fueled competition to build new service to nearly 20,000 unserved and underserved locations across the state. Over the past five years, we have been fortunate to receive Federal and state support for this work. I'm excited to see how this enthusiasm will grow as we develop new AI tools. In Nebraska, AI could become a tremendous support for our agricultural producers. Precision agriculture tools are fantastic ways to conserve water, monitor livestock health, and target application of fertilizers and pesticides, saving our farmers money and protecting the environment at the same time. These AI-powered tools are the future of agriculture. But we need reliable, universally available broadband in order to use them. As adoption of--and reliance on--AI technology increases, it will become even more important to ensure that all Americans have access to reliable, high-speed broadband networks. We need to ensure that rural Americans are not left behind as this technology moves forward. For the average citizen, AI technology is only useful if they have access to broadband connectivity. We must continue funding rural buildout, or this technology simply won't be available for all who need it. I also want to mention the importance of fiber buildout for an AI future. While we recognize that many technologies may be able to deliver broadband to our citizens, our commission has focused on fiber deployments for a few reasons: First, scalability. Not every technology can easily meet the FCC's current 100/20 Mbps threshold defining broadband-capable speed. Meanwhile, fiber-optic networks can greatly exceed those speed thresholds, and are capable of meeting our needs into the foreseeable future. I want to point out that data centers aren't using wireless or satellite technology to move data-- they are using fiber. Second, supporting development. Fiber is the backbone for basically all other technologies. We need fiber-based networks to support wireless, fixed wireless, and cable networks. Third, reliability. Buried fiber is much less susceptible to weather hazards and natural disasters. Fires, tornadoes, and ice storms in recent years in Nebraska have shown the need for the resiliency that fiber can provide. I also want to mention that deploying broadband infrastructure will not be enough if these networks are not maintained. We have heard from industry providers that although the investment in fiber itself may last for 40 years, the circuitry required by fiber must be upgraded much more frequently to accommodate changing technology. This is to be expected as demands on the networks for download and upload capacity, as well as latency demands, continue to grow. The tasks AI is capable of today will be eclipsed by what it can do five years from now--so long as the network is upgraded to keep up. For that reason, I want to emphasize that new buildout must be as future-proof as possible in the face of developing technology. Senator Fischer, I know that you recognize the importance of maintaining broadband networks. Hopefully it is clear that AI technology is dependent on these networks. I also think it is important to recognize that many of these networks are supported through Federal and state universal service funds. I, and the Nebraskans I represent, are grateful for your work and the work of your colleagues on the Universal Service Fund Working Group. I look forward to hearing the group's recommendations for the future of the USF. I know there is discussion as to how the Federal USF is funded. I would just point out that AI creates, and will create, significant demand on the network--a network that is supported today through surcharges on voice service. Thank you for the invitation to speak with you today. I'm happy to answer any questions you may have. Senator Fischer. Thank you, Commissioner. Next, we have Mr. Everson. You are recognized for five minutes. STATEMENT OF BOB EVERSON, CHIEF ARCHITECT OF PROVIDER MOBILITY, CISCO SYSTEMS, INC. Mr. Everson. Thank you. Chairman Fischer, Ranking Member Lujan, Chairman Cruz, Ranking Member Cantwell, and distinguished Senators, thank you for the invitation to testify. My name is Bob Everson. For 30 years at Cisco, I have worked on the systems powering our Nation's critical networks, and I have never seen a shift this consequential. Artificial intelligence is not just another application, it is driving a new connectivity fabric that will help define U.S. economic strength and national security for decades. Today, I will focus on how AI is transforming our networks and how Cisco is using AI to improve them. First, AI is changing not only the volume of network traffic, but the behavior. Cisco measured a fourfold increase in AI inference traffic over 8 months. Networks have traditionally been optimized for content flowing downstream, AI is far more two- way and uplink-intensive. Prompts, context, sensor data, and agent activity, all travel back toward AI models, and the resulting connections are active longer than conventional web transactions. AI agents will amplify these effects by operating at software speed. In our testing, an agent generated 450 percent more traffic than a person performing the same task, and roughly 70 percent of that additional traffic was inference. Meeting these demands requires more balanced network designs, greater upstream capacity, resilient paths with end-to-end visibility, stronger security, and quality of service that can be tailored to AI traffic. As models become faster, the locations where computing occurs, from centralized clouds to the edge, will increasingly affect performance, cost, and data governance. A network this fast and distributed cannot be operated solely at human pace. Through agentic ops, Cisco uses AI to support self-healing operations, identifying faults, rerouting traffic, and recovering at machine speed. Our deep network models can distinguish legitimate AI traffic from anomalies, help predict and isolate threats, identify vulnerabilities, and guide operators in deploying patches. AI-native mobile networks can use spectrum and infrastructure more efficiently, adapt in real time, and enable new services. For service providers, this can also turn the network into a platform for growth, opening new revenue streams beyond connectivity, through services like inference optimization, edge AI, trusted AI services, and industry- specific solutions. One noteworthy example is integrated sensing and communication, which allows AI to combine radio signals with vision and other inputs so wireless networks can connect devices while also detecting objects and movement, so support applications such as safer robotics, more efficient buildings, enhanced public safety, and even healthcare. Together, these capabilities define a new generation of intelligent, secure, and distributed networks that carry AI, use AI to improve operations, and enable AI-native applications and services, strengthening economic competitiveness and national security. I commend NTIA for this week's call to action for 6G leadership in security, which advances the international cooperation needed to ensure trusted partners shape the 6G future. To sustain U.S. leadership, I offer three policy priorities. First, accelerate the U.S. AI-native stack. Cisco is investing across multiple dimensions of AI-native networking, bringing new capabilities to 5G advance today while building the foundation for 6G. One example of this commitment is AI-WIN, a collaboration among Cisco, NVIDIA, MITRE, the O-RAN Development Company, Booz Allen, and T-Mobile, which brings AI, compute, networking, and wireless together to create a secure American-led path from 5G advance to AI-native 6G. I would encourage Congress to lean in on American strengths in these areas. Second, modernize permitting in infrastructure. As computing becomes more distributed, permitting must enable rapid and responsible deployment. As this committee considers the future of Universal Service Fund, it should account for the evolving costs of AI-ready networks so rural and urban communities can share in the benefits. Third, maintain a balanced spectrum policy. The 800 megahertz of licenses spectrum recently made available by Congress, is essential to high-capacity, high-uplink connectivity. The FCC's authorization of the 6 gigahertz ban for unlicensed use is equally important to meeting enterprise demand. A dependable pipeline of both is foundational to American leadership, and I thank you for your efforts to rebuild it. Cisco stands ready to work with this committee to ensure the U.S. remains the world's premier destination for AI and wireless innovation. Thank you for the opportunity to testify, and I look forward to your questions. [The prepared statement of Mr. Everson follows:] Prepared Statement of Bob Everson, Chief Architect, Provider Mobility at Cisco Systems, Inc. 1. Introduction. Chairman Fischer, Ranking Member Lujan, Chairman Cruz, Ranking Member Cantwell, and distinguished Senators, thank you for the invitation to testify today. My name is Bob Everson, and I am a proud resident of Dallas, Texas. I have spent more than 25 years at Cisco, where I currently serve as chief architect of our provider mobility team. In this role, I lead Cisco's platform definition and development, end-to-end architecture, ecosystem strategy, and strategic partnerships. My focus includes developing AI systems that operate across distributed environments and defining the mobile networking technologies, architectures, and products that advance enterprise mobility, IoT, and the platforms and ecosystems needed to bring these capabilities to scale. For decades, Cisco has served as the foundational layer for the world's most critical networks, supporting a diverse customer base that spans service providers, global enterprises, government agencies, and small-to-medium businesses. Today, Cisco is building the critical infrastructure for the AI era. Our goal is to ensure the United States remains the global leader in fixed and wireless technology by building a secure, interoperable infrastructure ecosystem where licensed, unlicensed, and satellite networks work in concert and where networks are secure, observable, and highly scalable. While Cisco develops the infrastructure, our customers--whether fixed or mobile service providers, enterprises, governments, or hyperscalers--operate their own networks. For service providers, Cisco provides the core infrastructure behind the tower and the Radio Access Network (RAN). Beyond hardware, we are integrating AI into the network itself. Cisco's deep expertise in network analytics and security makes it possible for us to automate operations, predict and prevent anomalies in real time, and enable customers to bring AI models into the enterprise with visibility and trust by leveraging the network. We recognize that AI workloads--characterized by massive data movement, high-density traffic, and some anticipated use cases requiring sub- millisecond latency--place unprecedented demands on the network. By combining our industry-leading networking, security, and observability portfolios, Cisco is ensuring that as organizations scale their AI initiatives, they do so on a foundation of resilience, security, and performance. I want to first thank this Committee, and Congress, for its tireless work to make 800 MHz of spectrum available for commercial use and for recognizing the importance of the 6 GHz band to enterprise networks powering schools, manufacturing facilities, hospitals, and campuses of all sizes. A balanced spectrum policy is critical to enhancing service for consumers and businesses in both rural and urban areas and across different modes and use cases of technology. I would also like to thank Federal Communications Commission (FCC) Chairman Brendan Carr and National Telecommunications and Information Administration (NTIA) Administrator Arielle Roth for their leadership in bringing these efforts to life. As I will explain, network traffic is only going to increase in volume and complexity, and these policies will help enable continued U.S. technological leadership. At Cisco, our approach toward AI in communications networks can be distilled into two questions: How is AI reshaping our networks, and how can networks leverage the power of AI? Today, I will address these questions and share how Cisco is building the infrastructure for more resilient, secure, and capable networks. My testimony draws on three recently released reports that draw from real-world traffic analysis, third-party data, Cisco-controlled lab tests with AI agents, and interviews with more than 6,000 organizations, all of which I have appended to my written testimony. 2. AI Usage Is Reshaping the Network. AI traffic has led to increased demand--we expect a 209 percent increase in campus and branch traffic volume in the next three years. However, the more consequential shift is in the nature of that traffic. Today, I will highlight what drives that change and how networks and policy must evolve to support its deployment and adoption.\1\ AI workloads are not like traditional web traffic: they will change traffic shape, symmetry, duration, and criticality.\2\ AI workloads are also twice as long in duration and significantly more upstream-heavy due to large, context-rich prompts. The use of new agentic AI, AI inferencing, and applications built on the connectivity fabric that AI- native networks will make possible are largely driving these workloads. --------------------------------------------------------------------------- \1\ ``AI Impact on Wide Area Networks,'' Cisco. (June 2026). Available at: https://www .cisco.com/c/dam/en/us/solutions/collateral/artificial-intelligence/ mass-scale-infrastructure/ai-network-traffic-report.pdf. \2\ Id. --------------------------------------------------------------------------- Agentic AI: AI agents are supercharging demand on the network, generating up to 450 percent more traffic per task than humans in a Cisco-controlled test. Unlike consumer-driven traffic, which produces minor, episodic spikes throughout the day, AI agents are always on.\3\ Agents may continually perform automated network configuration checks, analyze telemetry data, or interact with other agents to perform business functions. --------------------------------------------------------------------------- \3\ Supra, Note 1. --------------------------------------------------------------------------- Additionally, survey data shows 67 percent of respondents reporting an increase in ``east-west'' traffic.\4\,\5\ Modern AI models are rarely processed by a single server; they are distributed across clusters of GPUs and compute nodes. These nodes must constantly communicate with one another to share model parameters, gradients, and training data, which creates a massive amount of lateral traffic within a network. Traditional planning assumptions (e.g., burstiness, downlink dominance, human-paced interactions) will need to adapt to the new reality of AI and agentic AI traffic that lasts longer, demands more upstream capacity, and operates at software speed. Continuous measurement of AI-specific Key Performance Indicators (KPIs) and adaptive network architectures will be essential to sustain expectations for performance and user experience, particularly as network operators use agentic operations to assist with network management functions. --------------------------------------------------------------------------- \4\ ``East-West'' traffic refers to the movement of data between servers, devices, or nodes within a network, rather than moving into or out of the network. \5\ ``No Time to Wait: The Accelerating Impact of AI on Campus and Branch Networks,'' Cisco. (June 2026). Available at: https:// www.cisco.com/c/dam/m/en_us/solutions/networking/ai-impact-campus- branch-networks/documents/the-accelerating-impact-of-ai-on-campus-and- branch-networks.pdf. --------------------------------------------------------------------------- AI Inferencing: While AI inference traffic is currently negligible compared to dominant categories such as video streaming, observed growth rates are exceptional. Token-consumption data shows nearly 10x year-over-year growth, while some service provider measurements show around 4x growth in just the last eight months.\6\ Sustained growth at these rates signals that AI traffic will become a meaningful component of overall network traffic by 2035.\7\ AI inference paths will also become strategic network assets, requiring high levels of resilience, observability, and differentiated treatment (e.g., Quality of Service (QoS) and path security).\8\ --------------------------------------------------------------------------- \6\ Supra, Note 1. \7\ Id. \8\ Id. --------------------------------------------------------------------------- Enterprise Networks: In campus and branch networks--like the one that powers the Senate office building we are sitting in today--we have already seen customers report a 34 percent increase in traffic tied to AI workloads over the last 12 months, and they expect to see a 96 percent increase this coming year.\9\ Half of enterprise customers report that AI demand is concentrated on their Wi-Fi networks, and 73 percent of organizations already face or expect to face campus and branch capacity limitations within the next 24 months.\10\ This is largely because large majorities of organizations report increases in east-west traffic, latency-sensitive traffic, and continuous, automated AI traffic.\11\ While the large majority of AI to date has come from foundation models running on central infrastructure, we are seeing enterprises deploy more small language models, open-source models, and specialized models--such as vision and voice models--which can be distributed throughout the network. Each of these characteristics underscores the value of the FCC's forward-thinking decision in 2020 to authorize the full 6 GHz band for unlicensed Wi-Fi use. Indeed, wireless leaders deploying 6 GHz-capable Wi-Fi networks report that the band is solving their congestion issues, enabling high-bandwidth applications, and supporting AI workloads and applications. These developments are translating into concrete economic benefits: one study reports that the U.S. 6 GHz decision will generate $1.2 trillion in economic value per year by 2027.\12\ --------------------------------------------------------------------------- \9\ Supra, Note 5. \10\ Id. \11\ Id. \12\ Telecom Advisory Services, ``Assessing the Economic Value of Wi-Fi in the United States'' (September 2024), Available at: https:// www.teleadvs.com/assessing-the-economic-value-of-wi-fi-in-the-united- states/. --------------------------------------------------------------------------- Infrastructure: AI is driving the shift toward edge computing. Service providers must also consider ``AI-native'' traffic profiles for several reasons, including technical considerations, cost, and data sovereignty and security issues. Technical: Physical AI use cases such as robotics, autonomous vehicles, and industrial automation could require sub-millisecond decision-making. If an autonomous robot sends data to a central cloud and has to wait for a response, the round-trip latency could be too high for safe, real-time operation. Cost: AI operations generate massive amounts of data. For example, high-definition video analytics for public safety can generate terabytes of data daily. Backhauling that data to a central cloud is prohibitively expensive and creates massive network congestion. Data Sovereignty and Security: Enterprises and governments are increasingly concerned about data sovereignty and security. Many customers have security or regulatory concerns about moving sensitive information across the public Internet to a third-party cloud. AI operations at the edge of the network address each of these challenges. Latency is dramatically reduced by cutting the distance data needs to travel. Processing data locally similarly allows for transmission of only the result to the cloud, reducing bandwidth and cloud-related costs. And operating at the edge keeps sensitive data local, reducing the risk of interception or unauthorized access. 3. Networks Will Leverage AI for Increased Performance and Resiliency. While AI workloads present several challenges for network operators seeking to ensure seamless performance, reliability, and security, there is a tremendous opportunity to leverage AI to deliver new applications and better performance, infuse security into the fabric of the network, and manage the increased complexity. Agentic Operations: Agentic AI will change the nature of traffic on the network, but it will also provide network operators new tools to operate at machine speed and deliver greater performance, efficiency, and security. For service providers, whose networks are the backbone of the entire digital economy, AgenticOps--the shift from manual, human- led network management to an autonomous, AI-driven operational model overseen by humans--is not just a luxury; it is a necessity for managing the exponential growth in traffic and the stringent latency requirements of AI-native applications. Cisco is helping service providers navigate the transition to AgenticOps by providing the infrastructure, telemetry, and intelligence required to manage the massive complexity of modern networks. Cisco's products help service providers close ``visibility gaps'' in complex, distributed networks. Cisco's deep visibility tools allow AI agents to map the entire path of a data flow across public and private infrastructure to identify the root cause of latency or packet loss in seconds rather than hours. AgenticOps also allows the network to act as a self-healing system. Cisco's AI-native tools enable the network to reroute traffic, adjust capacity, or reconfigure network nodes when the system detects performance degradation or an impending hardware failure.\13\ This dramatically increases uptime and reliability for mission-critical services. AI-driven networks can also help deliver greater energy efficiency by continuously optimizing network operations through intelligent automation using real-time telemetry.\14\ AgenticOps enables networks to dynamically adjust power usage, traffic routing, and resource allocation, allowing operators to power down underutilized components during low-demand periods and reduce energy consumption without compromising service quality.\15\ --------------------------------------------------------------------------- \13\ Masum Mir, ``Leading the Next Era of Intelligent Connectivity,'' (Oct. 28, 2025). Available at: https://blogs.cisco.com/ news/leading-the-next-era-of-intelligent-connectivity?dtid=osolie00145 6&utm_loc=ar&utm_lang=en. \14\ ``Energy Management in the AI Era: Cisco's Digital Transformation Advantage,'' Available at: https://blogs.cisco.com/ industries/energy-management-in-the-ai-era-ciscos-digital-transforma tion-advantage \15\ Id. --------------------------------------------------------------------------- One of the most significant challenges for network operators is the talent gap in managing increasingly complex, software-defined networks. AgenticOps allows operators to automate repetitive, low-value tasks-- such as ticket resolution, configuration updates, and routine maintenance. These tools also help close the workforce talent gap by lowering the barrier to entry and allowing more junior analysts to ramp up quickly. By automating these tasks, Cisco's AI-enabled platforms can free network engineers to focus on higher-level architectural strategy and innovation, and free cybersecurity analysts to dedicate more time to strategic threat hunting and detection engineering. Resiliency and Security: In an AgenticOps environment, the network transitions from a passive transport layer to the primary cybersecurity enforcement point. Cisco empowers service providers to implement AI- native security that operates at machine speed--a critical requirement in today's dynamic threat landscape.\16\ By leveraging Cisco's purpose- built Deep Network Models, we analyze traffic patterns in real time to distinguish between legitimate AI-driven workloads and malicious anomalies.\17\ This predictive capability allows our customers to identify vulnerabilities before they escalate. Should a device or service be compromised, the network can autonomously isolate the threat, halt lateral movement and safeguard the broader service provider infrastructure against sophisticated cyberattacks.\18\ --------------------------------------------------------------------------- \16\ ``Shields Up: Cisco Live Protect Closes Vulnerability Gap with Compensating Controls,'' June, 2, 2026. Available at: https:// blogs.cisco.com/news/shields-up-cisco-live-protect-closes- vulnerability-gap-with-compensating-controls \17\ ``Cisco Hypershield--Our Vision to Combat Unknown Vulnerabilities,'' Available at: https://blogs.cisco.com/security/ cisco-hypershield-our-vision-to-combat-unknown-vulnerabilities. \18\ Id. --------------------------------------------------------------------------- Moreover, AI-native architectures enable network operators to simulate and validate cybersecurity patches on a digital twin of the network.\19\ This capability eliminates the ``patching paradox''--where the fear of downtime prevents critical security updates--by ensuring that every change is tested for stability and performance at machine speed. To maintain essential human oversight in an AgenticOps environment, we enable network operators to leverage a collaborative interface that allows security engineers to visualize, discuss, and refine complex policies with AI assistance.\20\ This human-in-the-loop approach ensures that while our network operations run at machine speed, strategic security decisions remain guided by human expertise and transparent, shared visibility. --------------------------------------------------------------------------- \19\ See, https://www.cisco.com/c/en/us/products/collateral/ security/hypershield/hypershield-so.html#Solvingrealcustomerchallenges. \20\ See, https://blogs.cisco.com/ai/ai-canvas-controlled- availability. --------------------------------------------------------------------------- AI-Enabled Network Applications: In addition to changes in traffic patterns, networks are moving toward AI-native platforms that will become the fabric of intelligent connectivity rather than a simple pipe. As network operators move compute toward the network edge--such as at a cell site where a tower sits--they will be able to run applications directly from the network. One promising application is Integrated Sensing and Communication (ISAC), which combines wireless communications and radio-frequency sensing to ``see'' objects' position and path using radio waves that reflect off them. Unlike optical sensors, it can detect intrusion even in low-light conditions, through smoke, or around obstructions where traditional video analytics might fail. This technology has been prototyped and demonstrated already, and it holds great promise for autonomous systems and robotics, AI-driven smart facilities, and public safety. Robotics: For example, ISAC enables the network to provide collision avoidance and path optimization, offloading some of the computational burden from the robot's own onboard sensors and ensuring safer operations in complex industrial environments. AI-Driven Smart Facilities: By sensing occupancy and environmental changes through ISAC, the network can dynamically adjust energy, HVAC, and lighting systems in enterprise campuses. This creates ``self-aware'' buildings that optimize for both human comfort and energy efficiency. Public Safety: Because the network can act as a pervasive radar system to track objects in restricted areas, ISAC could support Counter-Unmanned Aircraft Systems missions leveraging commercial networks around a stadium or critical infrastructure. Cisco and its partners are already exploring this technology, having demonstrated an early version at NVIDIA's GTC DC Conference last year and at Mobile World Congress Barcelona this spring. The network will also enable `Physical AI', which is the integration of digital intelligence with physical action that allows machines to operate independently in environments where conditions change continuously.\21\ While traditional AI is confined to software- based analysis, physical AI is designed to interact with the physical world. Physical AI applications--whether a warehouse robot rerouting around an unexpected obstacle, robotics in a manufacturing facility, or intelligent transportation infrastructure--will drive significant demand and constraints on the network while also opening new business monetization opportunities for network operators. --------------------------------------------------------------------------- \21\ See, https://www.cisco.com/site/us/en/learn/topics/artificial- intelligence/what-is-physical-ai.html. --------------------------------------------------------------------------- Demand is driven by several factors: 1) data collection from many types of sensors such as cameras for visual detail, lidar for depth, and radar for motion detection; 2) decision-making and reinforcement learning (where the model improves by attempting physical actions and receiving feedback to improve future action); and, 3) performing physical action through mechanical components like motors, conveyors, or robotic arms.\22\ --------------------------------------------------------------------------- \22\ Id. --------------------------------------------------------------------------- Each of these applications represents how AI is changing the network design: they require compute at the edge. They require an AI- native network core capable of integrating terrestrial and non- terrestrial networks and they require additional capacity to manage east-west traffic as agents, devices, and networks work in concert. Ultimately, Cisco envisions a combination of interoperable technologies to ensure ubiquitous, seamless, secure, and reliable connectivity for all end users, regardless of whether communications travel on wired, licensed, unlicensed, or satellite networks. 4. Policy Considerations. As this Committee considers the changing landscape, I offer three suggestions that will help enable continued U.S. leadership and better outcomes for American consumers and businesses. First, we must work together to protect American leadership in AI and compute and remain focused on building an American AI-native tech stack for wireless networks. However, we cannot wait for 6G standards to begin deployment; and we aren't. Through the AI-WIN partnership, Cisco, NVIDIA, MITRE, Booz Allen Hamilton, ODC, and T-Mobile are collaborating to develop America's first AI-native wireless telecommunications stack, while also developing innovative technologies to support new use cases in 5G-Advanced networks. We seek to leverage American leadership in AI and compute to demonstrate technical success through real-world deployments, ultimately shaping international standards to meet American companies where the technology is moving. I encourage Congress to lean in on areas where the United States has a strategic leadership role, such as compute, core networking, and applications. Second, as compute infrastructure becomes more distributed and moves closer to the network edge, service providers may face increased infrastructure demands. Policies that enable the responsible and efficient permitting and deployment of this infrastructure will be essential to helping service providers offer new services as the network becomes an AI-native platform. Additionally, as Congress considers the future of the Universal Service Fund, we encourage you to consider how these changes in network traffic will affect the operational costs borne by rural health care facilities, service providers, and schools, regardless of whether they are rural or urban. Finally, I encourage this Committee to continue its bipartisan approach to a balanced spectrum policy. As WRC-27 approaches, the United States needs to maintain its global leadership and expand its support for unlicensed operations in the 6 GHz band, including by authorizing higher-power operations and allowing expanded use in key enterprise environments. Likewise, the 800 MHz of spectrum made available for mobile use will be essential to enabling the connectivity needed to power new applications on AI-native networks, including the volume of uplink traffic we expect to see. 5. Conclusion. We are at an exciting inflection point in how AI is transforming our communications networks. As Cisco works with service providers, critical infrastructure operators, enterprises, and governments to deploy the foundational infrastructure of the AI era, we are proud to offer our expertise to this Committee and to Congress to help shape a technology future that connects consumers, businesses, and communities. Thank you for the invitation to testify, and I look forward to answering your questions. Senator Fischer. Thank you, Mr. Everson. Mr. Ramzanali, welcome. You are recognized for 5 minutes. STATEMENT OF ASAD RAMZANALI, DIRECTOR OF ARTIFICIAL INTELLIGENCE AND TECHNOLOGY POLICY, VANDERBILT POLICY ACCELERATOR, VANDERBILT UNIVERSITY Mr. Ramzanali. Chairman Fischer, Ranking Member Lujan, and Members of the Subcommittee, thank you for holding this important hearing. My name is Asad Ramzanali, and I am the Director of AI and Technology Policy at the Vanderbilt Policy Accelerator. AI and telecommunications are deeply intertwined. Fiber networks are how data moves in and out of AI data centers. And for Americans to benefit from AI, every American must have access to that system. We must reprioritize deploying high- speed, reliable, and symmetrical networks that reach every home in this country. Because I agree with much of what my co-panelists have said, I won't repeat them. Instead, I will focus--I will start with one specific threat that AI poses to our telecommunications systems, and that is cybersecurity. Well before AI, our telecommunications networks were vulnerable. A couple of years ago, the Chinese hacking group, Salt Typhoon, breached at least eight U.S. telecommunications providers, and it was actually able to intercept calls and text messages from high-profile targets, including Presidential candidates. One investigation called our systems a house full of open windows. But Salt Typhoon did it the old way, the low-tech, pre-AI way. In the last year, AI has started strengthening cyber attackers' methods. We have now seen several AI-enabled cyberattacks, including at least one that breached another country's telecommunications systems. As AI-enabled cyber threats started expanding, the FCC went the other way. It rescinded the cybersecurity rules that were put in place after the Salt Typhoon. That is the opposite of what we need to be doing. Now, we have even more reason for concern. Just last week, an open AI model, in testing, bypassed its sandbox constraints and hacked an AI--that AI platform Hugging Face. I recommend three actions on this front. First, require telecommunications providers to secure their networks, and similar requirements should be put in place for all critical infrastructure operators. Second, mandate pre-deployment testing of leading AI models. Testing should cover closed and open-weight systems alike, but the mitigations to reduce the identified risks will differ. This should not become a ban on open-weight models. Open-weight models are critical for competition, for research, and for cybersecurity. Third, this committee should open an investigation into the open AI Hugging Face incident. I will now use the remainder of my time to talk about how America's history with telecommunications, almost two centuries of history, can actually help inform the AI governance debates of today. Since the telegraph, communications providers have been required to treat similar customers alike. That is the principle of non-discrimination. As an example, consider that the FCC barred AT&T from blocking third-party equipment on its network, and that opened the door to early innovations like the answering machine, the fax machine, and even the modem that gave us the internet. AI models, or AI model providers are now gatekeepers in their own right, and they have their own conflicts of interest. Here is one example. Last summer, as Claude Code started becoming popular, Anthropic cutoff model access to an independent AI startup called Windsurf, that also did AI coding. Anthropic was Windsurf's critical supplier, and it became its competitor. That is the conflict of interest. Gatekeepers pick winners and losers instead of consumers, competition, and markets deciding outcomes. Two, shared infrastructure--shared jurisdiction. Telecommunications has always been governed by Federal, state, and local governments together. Today's data center buildout has no coherent governance strategy at all, and it is provoking real backlash around the country. Dozens of shelved projects, hundreds of protests, and hundreds of thousands of signatures just this year. With few policy levers at their disposal, hundreds of municipalities have turned to moratoriums, and one state has done the same. Instead of states subsidizing data center construction, Congress could establish a regime where local governments can charge a franchise fee so that they can compete for benefits, and they retain the right to say no. Finally, telecommunications is also instructive in the form of overspending on infrastructure that leads to bubbles. Today, trillions of dollars of investment are chasing tens of billions of dollars of AI revenue. This is more than the country has spent as a relative share of GDP than on the Manhattan Project, the Apollo program, the electricity buildout, or the interstate highway system. The financing mechanisms involved are also troubling. There is circular financing, which happened during the dot-com bubble. There is private credit, which Jamie Dimon has warned is full of cockroaches. There are special purpose vehicles, which Enron famously use, and there is asset-backed securities, which we all learned about in 2008. This committee investigated the financial engineering that enabled the dot-com bubble, but it did so after the bubble burst. In this instance, we can start those investigations now. There are myriad other things to discuss in the AI debate. I tried to focus on the infrastructural aspects of it. I thank you for holding this important hearing. And I look forward to your questions. [The prepared statement of Mr. Ramzanali follows:] Prepared Statement of Asad Ramzanali, Director of Artificial Intelligence and Technology Policy, Vanderbilt Policy Accelerator, Vanderbilt University Chairs Fischer and Cruz, Ranking Members Lujan and Cantwell, and members of the Subcommittee, thank you for holding this important hearing and inviting me to appear before you today. My name is Asad Ramzanali, and I'm the Director of Artificial Intelligence and Technology Policy at the Vanderbilt Policy Accelerator (VPA), a center at Vanderbilt University. I have experience working in the technology industry and in technology policy, including previously serving as the chief of staff and deputy director for strategy at the White House Office of Science and Technology Policy (OSTP), and as a staffer for members of the House and Senate. I started my career on the business side of Silicon Valley--first on the corporate strategy team at Intuit, then managing a startup accelerator backed by JPMorgan. My testimony reflects my personal views and does not represent the views of Vanderbilt University or VPA. Telecommunications and AI are deeply interdependent. At the most basic level, AI data centers depend on fiber networks to move data in and out of their facilities. More critically, American leadership in AI depends on all Americans being able to access the benefits of AI. This is why we need to deploy high-speed, reliable broadband to every home and business in the country. The biggest threat AI presents to telecommunications is cybersecurity. Network intrusions by Salt Typhoon showed that the telecommunications industry's cybersecurity practices are akin to ``a house full of open windows,'' as one investigation described it.\1\ Just when AI tools started showing a capacity to enable stronger cyberattacks, the Federal Communications Commission (FCC) weakened cybersecurity requirements. Headline-grabbing cyber incidents like OpenAI's recent model testing episode involving Hugging Face make abundantly clear that today's cybersecurity regime is no longer commensurate with the threat landscape. A near-term path forward has two tenets: First, telecommunications carriers should be required to harden cyber defenses. Second, the government should mandate pre- deployment model testing and require mitigations to identified risks. Those mitigations must account for the distinct affordances of closed and open-weight models. No testing requirement should result in a ban on open-weight models. --------------------------------------------------------------------------- \1\ Greg Otto, A House Full of Open Windows: Why Telecoms May Never Purge Their Networks of Salt Typhoon, CyberScoop (May 21, 2025), https://cyberscoop.com/salt-typhoon-chinese-hackers-us-telecom-breach/. --------------------------------------------------------------------------- Next, the long history of telecom policy offers concrete lessons for AI governance debates. Consider three facets of this history. First, since the earliest days of telecommunications, nearly two centuries ago, providers have operated under nondiscrimination regimes in which a provider must treat similarly situated customers the same. AI model providers have now also become gatekeepers and should be prohibited from picking winners and losers among startups and companies that build atop their models. Second, telecommunications policy is often governed under a shared federal-state-local policy regime. Federal law, for instance, lets local governments charge cable companies a franchise fee. The unprecedented construction of AI data centers, which is also drawing record levels of public outcry, could benefit from a similar coherent regime that clarifies rather than eliminates the roles of each level of government. Preserving local control should enable municipal-level data center moratoriums, but a minimum franchise fee would let communities negotiate to find real benefits. Third, after the dot-com bubble burst, this committee investigated the fraud and financial engineering that contributed to that bubble. Today, the global economy is in an AI overinvestment bubble of far larger magnitude and involving more complex financial engineering. Congressional investigations of those practices before today's bubble bursts would be far better than after. The lessons from telecom history are far from perfect analogs, but they at least provide some precedents on governing technologies. I. The Intersection of AI and Telecommunications A. How Telecommunications Networks Enable AI One common way to conceptualize the supply chain for AI is by illustrating it as a technological stack.\2\ In this model, consumers and businesses use (1) applications that are developed atop (2) AI models which are built (i.e., training) and accessed (i.e., inference) using (3) cloud computing infrastructure located in data centers full of (4) chips (i.e., semiconductors). Data centers are facilities that house the physical aspects of the AI Tech Stack (i.e., servers full of chips). Data centers depend on energy (electricity), cooling (water), and interconnect (fiber). Each layer depends on the technology and markets below. --------------------------------------------------------------------------- \2\ see, e.g., America's AI Action Plan, The White House, 20 (2025), https://www.white house.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf; Tejas N. Narechania & Ganesh Sitaraman, An Antimonopoly Approach to Governing Artificial Intelligence, 43 Yale L. & Pol'y Rev. 95, 108-128 (2024). This AI Tech Stack, like any model, is an oversimplification. Many other aspects of software, hardware, and networks could be added. And of course, AI depends on a broader socio-technical ecosystem that includes a deep base of research and development (R&D), prior technological advancements (e.g., the internet, semiconductors, smartphones, batteries), a well-trained workforce, a competitive marketplace, and significant intellectual and creative content, among other things. Telecommunications networks play two essential roles in enabling AI. First, data centers predominantly rely on fiber networks. The broader data center debate is often focused on other key inputs (e.g., electricity and water), in part because of particular constraints or issues with those inputs. For example, data centers' energy use and its effect on electricity prices are constantly in the headlines. The fact that fiber networks are not the focus of the policy debates should not be seen as reducing their role. Fiber networks are how data moves in and out of data centers, which explains why tier 1 data center cluster where fiber is readily available.\3\ --------------------------------------------------------------------------- \3\ Ritvik Chennupati, The Geography of AI: Where Are Data Centers Located?, Geo. U. Steers Ctr. For Glob. Real Assets (2025), https:// globalrealassets.georgetown.edu/insight/ai-driven-data-center- development/ (Tier 1 clusters [Northern Virginia, Atlanta, Silicon Valley, Dallas, Chicago, Phoenix, Hillsboro] formed near ``deep fiber routes and peering points,'' cheap power, experienced permitting ecosystems, and large land parcels near metro areas); Billy Roberts, Data Center Infrastructure in the United States, Nat'l Lab. of the Rockies (2025), https://research-hub.nlr.gov/en/publications/data- center-infrastructure-in-the-united-states-november-2025-2/. --------------------------------------------------------------------------- Second, the success of American AI depends on Americans being able to use AI tools, especially when their jobs or education depend on it. In that context, consumer access to high-speed, low-latency, reliable broadband is a critical policy objective that has been elusive for decades. Today, only 60 percent of homes are passed by fiber.\4\ Unfortunately, government support for closing this digital divide has made too little progress recently.\5\ Congress should re-prioritize deploying high-quality networks to every home and business in the country, along with deploying the necessary middle-mile and other support networks. --------------------------------------------------------------------------- \4\ Fiber Broadband Assoc. & Cartesian, Fiber Deployment Cost Annual Report, 4 (2025), https://fiberbroadband.org/resources/fiber- deployment-cost-annual-report-2025/. \5\ Allison Mollenkamp, Billions in Rural Broadband Funds Still on Hold, Roll Call, June 30, 2026, https://rollcall.com/2026/06/30/ billions-in-rural-broadband-funds-still-on-hold/ (Administration officials have also ``deferred specifics on the question of whether those funds will be held back from states that enact regulations on artificial intelligence, as directed by a presidential executive order.''). --------------------------------------------------------------------------- B. AI-Enabled Opportunities and Threats to Telecommunications Networks The opportunities of AI adoption by the telecommunication ecosystem come in many forms. I'll defer to other witnesses to speak to how industry actors are using AI to increase efficiency and in other ways. I'll note two uses that have specific public policy effects: When a caller dials 9-1-1 speaking a language other than English, a public safety answering point (PSAP) (i.e., 9-1-1 call center) can take minutes to identify which language the caller is speaking to transfer them to the appropriate responder. AI-based systems being used today can identify the caller's language and transfer the emergency call to the appropriate person in seconds. PSAPs are also using AI to transcribe calls, reduce call background noise, and in other ways.\6\ --------------------------------------------------------------------------- \6\ NGA 911, How AI Is Changing 911 Centers: From Non-Emergency Overload to Smarter Emergency Response, (July 15, 2026), https:// nga911.com/press-releases/post/how-ai-changing-911-centers-non- emergency-overload-smarter-emergency-response; Nat'l Telecomm. & Info. Admin., Off. Of Pub. Safety Commc'ns, Improving 9-1-1 Operations with Artificial Intelligence, https://www.ntia.gov/sites/default/files/2025- 11/ntia-ai-in-9-1-1-white-paper.pdf (last visited July 29, 2026); Cybersec. & Infrastructure Sec. Agency SAFECOM, Artificial Intelligence in Emergency Communications Centers (2026), https://www.cisa.gov/sites/ default/files/2026-04/2026.01.16_s-n_ai-implemen- ecc_infographic_508C_FINAL.pdf. Given the limited nature of radiofrequency spectrum available for the wide array of domains where it could be used for commercial, scientific, national security, and other objectives, AI continues to hold promise in enabling more effective spectrum sharing and dynamic spectrum allocation. Late last year, the Pentagon announced a demonstration project with five industry and academic teams as part of the Advanced Spectrum Coexistence project, mandated under the 2023 National Spectrum Strategy.\7\ --------------------------------------------------------------------------- \7\ Lisbeth Perez, DOD Partners With Industry, Academia on Spectrum Sharing, MeriTalk, Oct. 6, 2025, https://www.meritalk.com/articles/dod- partners-with-industry-academia-on-spectrum-sharing/. In contrast to those opportunities, the advancement of AI also exacerbates one major threat to telecommunications networks: cybersecurity. Before AI started being used for cyberattacks at scale, telecommunications providers were vulnerable to intrusions and experienced major ones. Starting in 2021, Salt Typhoon, a Chinese state-sponsored group, breached at least eight U.S. carriers. It reached the systems that carriers use to fulfill lawful-intercept requests, allowing it to geolocate users and intercept the calls and texts of high-profile targets, including then-candidates Trump, Vance, and Harris. This attack was carried out largely by using stolen credentials and exploiting known, unpatched equipment vulnerabilities; more advanced hacking techniques, like exploiting zero-day vulnerabilities, have not been reported as part of this event.\8\ One press investigation called the U.S. telecommunications system ``a house full of open windows,'' arguing the carriers may ``never fully expunge the hacking group from their networks.'' \9\ As recently as February of this year, an FBI official confirmed the threat posed to U.S. telecommunications networks from Salt Typhoon ``is still very, very much ongoing.'' \10\ I appreciate Ranking Member Cantwell's oversight efforts to highlight the critical nature of these attacks.\11\ --------------------------------------------------------------------------- \8\ Nat'l Sec. Agency, et al.,Countering Chinese State-Sponsored Actors Compromise of Networks Worldwide to Feed Global Espionage System (2025), https://media.defense.gov/2025/Aug/22/2003786665/-1/-1/0/ CSA_COUNTERING_CHINA_STATE_ACTORS_COMPROMISE_OF_NETWORKS.PDF; Sarah Krouse, Robert McMillan & Dustin Volz, China-Linked Hackers Breach U.S. Internet Providers in New `Salt Typhoon' Cyberattack, Wall Street Journal, Sept. 25, 2024, https://www.wsj.com/politics/national- security/china-cyberattack-internet-providers-260bd835. \9\ Otto, supra note 1. \10\ Derek B. Johnson, FBI: Threats from Salt Typhoon Are `Still Very Much Ongoing,' CyberScoop (Feb. 19, 2026), https://cyberscoop.com/ fbi-salt-typhoon-ongoing-threat-cybertalks-2026/. \11\ Derek B. Johnson, Cantwell Claims Telecoms Blocked Release of Salt Typhoon Report, CyberScoop (Feb. 3, 2026), https://cyberscoop.com/ cantwell-claims-telecoms-blocked-release-of-salt-typhoon-report/. --------------------------------------------------------------------------- In response to the Salt Typhoon intrusion, in January 2025, the FCC under Chairwoman Jessica Rosenworcel required carriers to secure their networks against cyber intrusion and also proposed mandatory cybersecurity risk-management plans.\12\ On November 20, 2025, the FCC under Chairman Brendan Carr reversed these efforts, relying instead on carriers' voluntary ``agreement to take extensive steps to protect national security.'' \13\ --------------------------------------------------------------------------- \12\ Protecting the Nation's Communications Systems from Cybersecurity Threats, Declaratory Ruling and Notice of Proposed Rulemaking, 40 FCC Rcd. 876 (2025). \13\ Quote from Fact Sheet in Protecting the Nation's Communications Systems from Cybersecurity Threats, Order on Reconsideration (as circulated on October 30, 2025), available at https://docs.fcc.gov/public/attachments/DOC-415190A1.pdf; the item was adopted and published as Protecting the Nation's Communications Systems from Cybersecurity Threats, Order on Reconsideration, 40 FCC Rcd. 9582 (2025). --------------------------------------------------------------------------- Even before the FCC weakened cybersecurity requirements, AI was strengthening the hand of cyber attackers. In August 2025, Anthropic reported that a Chinese threat actor successfully used Claude to enhance a cyber offensive that ultimately appears to have compromised Vietnamese telecommunications providers and government databases.\14\ Days before the FCC's reversal, Anthropic disclosed it had disrupted what it called the first documented large-scale AI-enabled cyberattack: a Chinese state-linked group manipulated Claude Code into automating most aspects of an espionage campaign against 30 targets.\15\ --------------------------------------------------------------------------- \14\ Anthropic, Threat Intelligence Report: August 2025, 18 (2025), https://www-cdn.anthropic .com/b2a76c6f6992465c09a6f2fce282f6c0cea8c200.pdf. \15\ Meaghan Tobin & Cade Metz, Anthropic Says Chinese Hackers Used Its A.I. in Online Attack, New York Times, Nov. 14, 2025, https:// www.nytimes.com/2025/11/14/business/chinese-hackers-artificial- intelligence.html. --------------------------------------------------------------------------- Earlier this month, an even starker case occurred. During OpenAI's internal testing, a model with safeguards deliberately reduced, bypassed the company's sandbox controls, gained Internet access, and breached the digital infrastructure of the AI-hosting platform Hugging Face, all to obtain the answers to the evaluation it was being run against. OpenAI later disclosed that its model breached at least four additional third-party accounts.\16\ As others have pointed out, OpenAI clearly should have been monitoring the incident more actively, could have used air gap systems for the test, and could have taken other steps to mitigate the effects of the test.\17\ This cyber incident requires serious study. This committee should launch an investigation to establish the facts and missteps. --------------------------------------------------------------------------- \16\ Dell Cameron & Maxwell Zeff, OpenAI's Rogue AI Agent Hacked More Than Just Hugging Face, Wired, July 2026, https://www.wired.com/ story/openais-rogue-ai-agent-hacked-more-than-just-hugging-face/. \17\ Harry Booth, How OpenAI Lost Control of an AI Model--and What Needs to Change, Time, July 24, 2026, https://time.com/article/2026/07/ 24/openai-hugging-face-attack/. --------------------------------------------------------------------------- However, gathering facts is necessary but insufficient. We are undergoing a meaningful escalation of cyber risks that requires a significant shift in policy thinking from two angles. First, telecommunications carriers, like all critical infrastructure operators and large organizations, should be required to harden their cyber defenses. In telecommunications, this starts with requiring carriers to secure their networks, as the FCC started doing in January 2025. Relatedly, as the telecommunications ecosystem transitions from largely analog equipment to software-defined systems, vendors should prioritize security-by-design, along with openness and interoperability. As a general principle, cybersecurity requirements should match the threat landscape such that critical infrastructure is resilient to state-of- the-art cyberattacks. Second, our posture toward cybersecurity must change. Today's regime for cybersecurity consists of companies voluntarily testing, red-teaming, and disclosing incidents; government agencies publishing best practices and working with companies in bespoke arrangements; and researchers responsibly disclosing vulnerabilities. This regime is insufficient for today's realities. The seriousness of AI accelerated risks from cybersecurity, and those associated with chemical, biological, radiological, and nuclear (CBRN) weapons, necessitates mandatory pre-deployment testing by government agencies that can work with operators of critical infrastructure and digital infrastructure companies to remediate vulnerabilities. Critically, testing requirements should apply to closed-weight and open-weight models alike, but the methods of mitigating risks identified in that testing will need to be distinct to account for the differing technical affordances of closed-weight and open-weight systems. This regime should not have the effect of banning open-weight models because of the critical role those play in improving competition, enabling research, and responding to cyber intrusions.\18\ Recently, in two separate letters, over 150 little tech startups\19\ and several large incumbents made these points.\20\ --------------------------------------------------------------------------- \18\ Kai Nicol-Schwarz, How a Chinese AI Model Stopped OpenAI's `Unprecedented' Cyber Attack, CNBC, July 24, 2026, https:// www.cnbc.com/2026/07/24/chinese-ai-model-openai-cyber-attack.html. \19\ Sophia Cai, Startup Founders Urge Trump Not to Shut off Chinese Open Weight AI, Politico, July 22, 2026, https:// www.politico.com/news/2026/07/22/startup-founders-urge-trump-not-to- shut-off-chinese-open-weight-ai-01008992. \20\ Katherine Long, Big Tech Companies Defend Open-Weight AI Models, Politco, July 24, 2026, https://www.politico.com/news/2026/07/ 24/big-tech-companies-defend-open-weight-ai-models-01010981. --------------------------------------------------------------------------- II. Lessons From Telecommunications Policy to Inform AI Governance Debates Beyond these specific areas of intersection between AI and telecommunications networks, the government's 180-year history of regulating the telecommunications sector can be instructive in how policymakers should approach AI governance. A. Nondiscrimination Rules Can Enable Innovation Nondiscrimination principles--treating similarly situated customers alike--have a long history and trace back to at least the Middle Ages.\21\ In the context of telecommunications, every major shift in communication technologies has been closely followed by applying nondiscrimination principles. In 1844, Samuel Morse transmitted the first message via telegraph on a demonstration line that Congress funded. In 1846, when Morse set up a corporation to commercialize his invention, the corporate bylaws of the organization, established under a specific chartering statute by the Maryland state legislature, included a provision specifying that, with respect to transmission of telegraph dispatches, ``the first to come shall be first served, without regard to sex, wealth, or station,'' subject to certain exceptions, including prioritizing government transmission in times of ``great public emergencies.'' \22\ This provision was inspired by the long-standing operating principles of the Post Office, which ultimately took over administration of the 1844 demonstration line, along with by a desire to curtail emergent speculative financial practices that capitalized on information speed.\23\ The idea was ported into state laws, starting with New York in 1848.\24\ In 1860, Congress instituted the first-in-first-out telegraph principle as a condition of the government-subsidized transcontinental telegraph line.\25\ --------------------------------------------------------------------------- \21\ Morgan Ricks et al., Networks, Platforms, and Utilities: Law and Policy 24 (2022). \22\ Articles of Association, and Charter from the State of Maryland, of The Magnetic Telegraph Company, Together with the Office Regulations and the Minutes of the Meetings of Stockholders and Board of Directors 3-4 (1847), available at https:// brittlebooks.library.illinois.edu/brittlebooks_open/Books2009-11/ magnte0001artass/magnte0001 artass.pdf. \23\ Richard R. John, Network Nation: Inventing American Telecommunications 77 (2010). \24\ An Act to Provide for the Incorporation and Regulation of Telegraph Companies, ch. 265, Sec. 12, 1848 N.Y. Laws 392, 395. \25\ Pacific Telegraph Act of 1860, 12 Stat. 41 (Chapter 137). --------------------------------------------------------------------------- A similar story played out with the telephone networks, though in different ways. Alexander Graham Bell invented the telephone in 1876. Three years later, the Connecticut legislature passed a law amending its telegraph statutes by simply adding ``or telephone'' after each reference to telegraph,\26\ which attached the state's existing nondiscrimination provision to telephone networks.\27\ In 1887, a court applied nondiscrimination requirements, as an aspect of common carrier requirements, to the telephone network. For decades, AT&T fought against subscribers using third-party equipment attached to its network.\28\ After multiple court cases and FCC proceedings on the matter, the Commission's landmark Carterfone decision in 1968 made clear that AT&T could not discriminate against third-party equipment.\29\ This decision is seen as creating the conditions for innovations like the answering machine, fax machine, and modem, which enabled the Internet itself.\30\ --------------------------------------------------------------------------- \26\ An Act relating to Telephone Companies, 1879, ch. 36, 1879 Conn. Pub. Acts 381. \27\ Conn. Gen. Stat. ch. 2, pt. 10, Sec. 8, at 342 (1875) (``Telegraph or telephone companies shall receive dispatches for any person and for other telegraph or telephone lines, and shall transmit them in the order of time in which they are received. . .'' emphasis added for changes to the statute). \28\ Ricks et al., supra note 20 at 339-340. \29\ For a discussion of how judges have leveraged nondiscrimination and other public utility remedies to Carterfone and other antitrust cases, see, Ganesh Sitaraman, The Public Utility Tradition in Antitrust Remedies, 75 Emory L.J. 1217 (2026). \30\ Michael Hoeker, From Carterfone to the iPhone: Consumer Choice in the Wireless Telecommunications Marketplace, 17 Commlaw Conspectus 187 (2008). --------------------------------------------------------------------------- AI model providers (e.g., OpenAI, Anthropic, Google) have business conflicts of interest that mean a nondiscrimination principle is worthy of consideration.\31\ As one example, consider that Anthropic curtailed Windsurf's API access to its leading Claude models because press reports suggested that OpenAI was slated to acquire the startup.\32\ Windsurf was then an independent AI coding application that was built atop Anthropic's models, among others. However, at the same time, Anthropic's own Claude Code was rising in popularity, creating an inescapable conflict: Anthropic was competing with its customer.\33\ Any provider that both supplies infrastructure to startups and competes with them faces this conflict, and a nondiscrimination rule should apply to all of them prospectively, regardless of which company happens to be ahead. In recent batches of startups selected by the accelerator Y Combinator, as many as 90 percent of startups were AI companies.\34\ An AI model provider has the capacity to pick winners and losers, rather than letting competitive forces, customer demand, and market force determine outcomes.\35\ For this reason, Congress should consider prohibiting AI model providers from unreasonably or unjustly discriminating via pricing, speed, or service quality among similarly situated customers.\36\ --------------------------------------------------------------------------- \31\ Asad Ramzanali & Akhil Rajan, AI Neutrality, Vanderbilt Policy Accelerator (2026), https://cdn.vanderbilt.edu/vu-URL/wp-content/ uploads/sites/412/2026/01/28222934/AI-Neutrality-.pdf. \32\ Maxwell Zeff, Anthropic Co-Founder on Cutting Access to Windsurf: ``It Would Be Odd for Us to Sell Claude to OpenAI,'' TechCrunch (June 5, 2025), https://techcrunch.com/2025/06/05/anthropic- co-founder-on-cutting-access-to-windsurf-it-would-be-odd-for-us-to- sell-claude-to-openai/. \33\ Mark MacCarthy, What Happens When AI Companies Compete with Their Customers?, Brookings (2026), https://www.brookings.edu/articles/ what-happens-when-ai-companies-compete-with-their-customers/. \34\ 90 Percent of Y Combinator Startups Are Now AI Companies-- Here's Why That's Alarming, Inc, Mar. 4, 2026, https://www.inc.com/inc- video/90-percent-of-y-combinator-startups-are-now-ai-companies-heres- why-thats-alarming-2/91312218. \35\ Asad Ramzanali & Tom Wheeler, To Unleash AI Innovation, Stop Model Providers from Picking the Winners, Brookings (May 18, 2026), https://www.brookings.edu/articles/to-unleash-ai-innovation-stop-model- providers-from-picking-the-winners/. \36\ For a full legislative recommendation, see Ramzanali and Rajan, supra note 30. --------------------------------------------------------------------------- B. Federal, State, and Local Governments Each Play Critical Roles in Governance Telecommunications networks have long been governed in a shared jurisdiction between federal, state, and local governments. From the earliest days of the telegraph to modern broadband buildouts, Congress and courts have recognized the importance of respecting this shared jurisdiction model. As one example, consider the Cable Communications Policy Act of 1984, which authorized a state or local government to grant cable franchises: reserve channel capacity for public, educational, or governmental (PEG) use; and charge a franchise fee up to 5 percent of cable revenues. It also authorized the FCC to set technical standards, issue regulations to remedy violations and promote diversity of information sources, and waive certain regulatory limits.\37\ --------------------------------------------------------------------------- \37\ Cable Communications Policy Act of 1984, Pub. L. No. 98-549, 98 Stat. 2779 (codified as amended in scattered sections of 47 U.S.C.). --------------------------------------------------------------------------- Today's data center buildout is happening without a coherent governance regime. The speed of the ungoverned buildout has been hit with serious backlash that includes dozens of shelved construction plans,\38\ hundreds of protests,\39\ and hundreds of thousands of petition signatures.\40\ People are understandably upset about everything from rising electricity costs to the effect construction has on water sources, to how constant noise affects nearby residents. Two underappreciated factors worsen these dynamics. First, the details of many data center deals are shielded behind nondisclosure agreements.\41\ Second, in at least 37 states, taxpayers are funding this construction through tax breaks and subsidies. In four states-- Georgia, Texas, Virginia, and Ohio--those breaks cost taxpayers over $1 billion per year.\42\ Often the bar for receiving those benefits is astonishingly low: In Missouri, a data center needs to create only 10 jobs and invest just $25 million to avoid all state and local sales and use taxes.\43\ Responses to these issues have been small relative to the volume of protest. Federal policymakers have proposed voluntary commitments for data centers to absorb energy price increases,\44\ and some states are considering rolling back their tax subsidies. Over 100 municipalities, left with few tools to respond, have enacted data center moratoriums, as has the state of New York. Moratoriums are inherently a blunt tool, provide no governance over the thousands of existing and planned data centers,\45\ and often lack clear factors for ending the pause. --------------------------------------------------------------------------- \38\ Simmone Shah, Community Backlash to AI Data Centers Is Growing Across the U.S., Time, July 2026, https://time.com/article/2026/07/22/ community-backlash-ai-data-centers/. \39\ Id. \40\ Danica Sauter & Kassidy Brown, Nashville Zoo Petition against Data Center Tops 500K Signatures, WSMV, June 21, 2026, https:// www.wsmv.com/2026/06/21/nashville-zoo-petition-against-data-center- tops-500k-signatures/. \41\ Deanna Noel, The Secret Data Center Buildout, Public Citizen (July 8, 2026), https://www.citizen.org/news/the-secret-data-center- buildout-how-states-can-stop-big-techs-abuse-of-ndas/. \42\ Kasia Tarczynska, Even Cloudier with a Greater Loss of Spending Control: How Data Center Tax Abatements Undermine Public Budgets, Good Jobs First (2026), https://goodjobsfirst.org/even- cloudier-with-a-greater-loss-of-spending-control-how-data-center-tax- abatements-undermine-public-budgets/. \43\ Mo. Dep't of Econ. Dev., Data Center Sales Tax Exemption Program Guidelines, available at https://ded.mo.gov/media/pdf/data- center-sales-tax-exemption-program-guidelines. \44\ Ratepayer Protection Pledge, The White House, https:// www.whitehouse.gov/ratepayer-protection-pledge/ (last visited July 29, 2026). \45\ Nicol Turner Lee & Darrell M. West, Data Center Moratoriums Are Not a Substitute for Oversight, Brookings (July 28, 2026), https:// www.brookings.edu/articles/data-center-moratoriums-are-not-a- substitute-for-oversight/. --------------------------------------------------------------------------- The cable regulation regime is far from perfect, and its political history is markedly different from today's data center debates. However, Congress could establish a real shared-jurisdiction data center regime that scraps state data center subsidies in favor of a minimum franchise fee paid to local governments willing to accept a data center, encouraging companies to bid up local benefits. Federal standards could require limits on emissions and encourage low-water cooling systems, among other things. Critically, such a regime should preserve a local jurisdiction's right to enact a moratorium since there's no need for every municipality in the country to accept data centers. C. Infrastructure Overinvestment Causes Bubbles Last but certainly not least, telecom history also warns about what happens when infrastructure investment outpaces revenues. Nearly daily, experts across government, finance, and even the technology industry itself point to the fact that the U.S. and global economy are in fact in a bubble based on overinvestment in AI.\46\ At core is a relatively simple math problem: trillions in investment are chasing tens of billions in revenue, which is agnostic of the pace of technological advancement. This year alone, hyperscalers are on track to invest more in capital expenditures relative to GDP than the U.S. experienced during the Manhattan Project, the expansion of electricity, the Apollo space program, or the construction of the interstate highway system.\47\ This is worsened by the fact that the investments are dominating nearly every capital and asset class, which increasingly includes circular financing (which enabled fraud in the dot-com bubble\48\), private credit (which Jamie Dimon recently compared to cockroaches\49\), special-purpose vehicles (which Enron famously used\50\), and credit default swaps and asset-backed securities (which contributed to the 2008 financial crash\51\).\52\ --------------------------------------------------------------------------- \46\ Examples from just this week include: Alexa LoMonaco, Jim Cramer Warns AI's Circular Financing Frenzy Echoes the Dot-Com Bubble, CNBC, July 27, 2026, https://www.cnbc.com/2026/07/27/jim-cramer-warns- ai-circular-financing-echoes-dot-com-bubble.html; John Stepek, Lessons From History About Today's AI Bubble, Bloomberg, July 28, 2026, https:/ /www.bloomberg.com/news/newsletters/2026-07-28/lessons-from-history- about-today-s-ai-bubble; Kathryn Carlson, We Might Be in an AI Bubble, Global Finance Watchdog Chief Says, Politico, July 28, 2026, https:// www.politico.eu/article/ai-bubble-financial-stability-board-john-schind ler/. \47\ Meghan Bobrowsky, Drew An-Pham & Alana Pipe, Big Tech's AI Push Is Costing a Lot More Than the Moon Landing, Wall Street Journal (Feb. 7, 2026), https://www.wsj.com/tech/ai/ai-spending-tech-companies- compared-02b90046 \48\ Capacity Swaps by Global Crossing and Qwest: Sham Transactions Designed To Boost Revenues?, H. Hrg. 107-129 before the Comm. on Energy & Com., 107th Cong. (2002); Tomasz Tunguz, Circular Financing: Does Nvidia's $110B Bet Echo the Telecom Bubble? (Oct. 3, 2025), https:// www.tomtunguz.com/nvidia_nortel_vendor_financing_comparison/. \49\ Elisabeth Buchwald & Matt Egan, Why Jamie Dimon Is Warning of `Cockroaches' in the U.S. Economy, CNN (Oct. 16, 2025), https:// www.cnn.com/2025/10/16/business/jamie-dimon-us-economy-cockroaches. \50\ Aisha Down & Dan Milmo, Nvidia Insists It Isn't Enron, but Its AI Deals Are Testing Investor Faith, The Guardian (Dec. 28, 2025), https://www.theguardian.com/technology/2025/dec/28/nvidia-insists-it- isnt-enron-but-its-ai-deals-are-testing-investor-faith \51\ Roge Karma, Something Ominous Is Happening in the AI Economy, The Atlantic (Dec. 10, 2025), https://www.theatlantic.com/economy/2025/ 12/nvidia-ai-financing-deals/685197. \52\ For a more detailed analysis, see, Asad Ramzanali, After the AI Crash, Vanderbilt Policy Accelerator (2026), https:// cdn.vanderbilt.edu/vu-URL/wp-content/uploads/sites/412/2026/03/ 23144242/After-the-AI-Crash.pdf. --------------------------------------------------------------------------- During the dot-com bubble that burst in the early 2000s, hundreds of thousands of Americans lost their jobs and the U.S. economy lost trillions in value. The Nasdaq lost nearly 80 percent of its value between its March 2000 high and October 2002 low. That bubble inflating was based on unproven tech business models, but its deeper causes involved an overinvestment in telecommunications infrastructure, complex financial engineering, and fraud.\53\ This committee held several hearings investigating those practices in 2001 and 2002.\54\ Of course, it would have been better to investigate the industry's troubling finances before the bubble burst. --------------------------------------------------------------------------- \53\ Id. \54\ See, e.g., Financial Turmoil in the Telecommunications Marketplace; Maintaining the Operations of Essential Communications Facilities, S. Hrg. 107-1136 before the S. Comm. on Com., Sci., & Transp., 117th Cong. (2002). --------------------------------------------------------------------------- As bad as the dot-com bubble bursting was, today's AI overinvestment is actually more like the even more economically calamitous 2008 financial crash because of the magnitude of the investments involved and the complex interlocking financing described. This committee should investigate these practices now, before a crash. Conclusion Telecommunications networks enable AI, and AI is changing aspects of telecommunications. The biggest change is the risk posed by AI- enabled cyberattacks. As Congress engages in various aspects of AI governance, telecommunications history presents many lessons and models worth considering. These are, of course, imperfect analogs, but they show that policymakers can and should enact policies to govern the risks of technologies. Thank you for inviting me to testify, and I look forward to your questions. Senator Fischer. Thank you, sir. I would like to recognize Senator Lujan for his questions at this time. He is bouncing back and forth with different obligations this morning. So Senator Lujan, you are recognized. Senator Lujan. Madam Chair, thank you so very much for that. And to all of our panelists, thank you for being here, and everyone that is tuning in. I will start with a simple yes or no question. Mr. Spalter, can AI be transformative for the people in America if they do not have access to broadband networks? Mr. Spalter. No. Senator Lujan. Commissioner? Mr. Watermeier. No. Senator Lujan. Mr. Everson? Mr. Everson. No. Senator Lujan. Mr. Ramzanali? Mr. Ramzanali. No. Senator Lujan. The $42.5 billion BEAD funding appropriated by Congress has been called a once-in-a-lifetime investment in connectivity, in reliable high-speed access across the country. Yet the administration has altered policies in a way that steer investments away from fiber buildout, and in doing so, have injected significant delays into the process. Denying states and territories the full amount of money appropriated for BEAD goes against the goal of connecting 100 percent of Americans across America. Mr. Spalter, can you explain how the BEAD restructuring notice impacted your member companies, and what that means for our ability to support AI applications at scale? Mr. Spalter. It has, in fact, made us all focus our minds on the additional levers that we have as an industry to accelerate our ability to not only build, but invest in, and expanding fiber connectivity to where it is needed most, and particularly to the unserved and underserved who stand to benefit significantly from AI connectivity. Which is why, though mindful of, and dutiful to the rules that have been established, and working closely with state broadband offices, we are imploring Congress, and other parts of the government, to accelerate our ability to build through things like permitting reform, and we support your initiative to do that. Senator Lujan. Mr. Spalter, what good is permitting reform if the money is taken away from you? What good is the money for your companies if permitting reform is passed, if the administration takes the money from you? Mr. Spalter. Well, where can---- Senator Lujan. We can get everything out of the way, but if you don't have any money, what does it matter? Mr. Spalter. Yes, our core focus as an industry is to do whatever we can to build fiber network connectivity as far as possible. We do so mightily through private investment and through public partnerships as much as we possibly can. Senator Lujan. Let me ask the question this way. If your member companies have full access to the $42 billion, will they be able to build more fiber? Mr. Spalter. Well, obviously, we will be able to. Senator Lujan. Yes? I appreciate that. Commissioner? Mr. Watermeier. Complicated question for me. BEAD was originally put in the Nebraska Public Service Commission, with the change in administration, there was a creation of the broadband office in Nebraska. I understand what happened inside of the world as far as the applications, different things. After the first application, many of our telecoms in the State of Nebraska had spent a great deal of money on engineering, and then was turned back, and had to reapply, they decided not to apply. So the long story short is, in my opinion, not from the Commission's opinion, but if there were--dollars were spent there, we could get it and faster in hands as far as fiber. Senator Lujan. I appreciate that, Commissioner. I learned from Chair Fischer that I have a few things in common with you. I was a Member of the Commission in New Mexico, also a farmer, so I already like you, Commissioner, so look forward to learning from you as well. And Mr. Everson, same question. Mr. Everson. So as an engineer, I am probably the least qualified up here to talk about this. I will say from our standpoint, you know, this is a once-in-a-generation opportunity with the funding, and getting everybody connected is really an essential aspect for us. Senator Lujan. I appreciate that. And Mr. Ramzanali, the United States considers how AI strengthens network performance and resilience, we need to equally be rigorous around the security, which I appreciate your attention to investigations and the challenge to us here in the Senate and in the Congress, I am in that space. I am concerned that there are many decisions that are being made moving toward a voluntary evaluation framework for these high- powered models, leaving the security of our public infrastructure dependent on the goodwill of individual developers. In your view, what role should government play in evaluating the capabilities of these high-powered models, especially when public infrastructure is at risk based on the choices of a small number of powerful and influential developers? Mr. Ramzanali. You asked the question the perfect way. We should not be dependent on the goodwill of private companies. We should be the ones who are deciding in public through their representatives and the agencies that Congress creates, should be the ones deciding what the standards are, should be helping making sure that critical infrastructure providers are ready when sophisticated AI models are released, and we should be the ones deciding when all of that happens. Senator Lujan. I appreciate that. Madam Chair, thank you for the flexibility and allowing me to ask questions first. I look forward to the rest of your responses, and look forward to working with each and every one of you. Thank you again, Madam Chair. Senator Fischer. Thank you, Senator Lujan. Nebraska is home to dozens of growing industries that are increasingly reliant on AI in their day-to-day operations. That is from precision agriculture, which, Commissioner, you talked about, manufacturing, logistics, freight, health care, financial services, public safety. And Nebraska businesses and their customers are integrating more advanced technologies like AI. They have to be resilient. You know, if we are looking at the applications, we want to make sure they are resilient. They have to be able to have that connectivity, which you talked about. Can you tell me how Nebraska's communications infrastructure, Commissioner, is ready for the complexity and volume of traffic that AI has introduced into the equation? Mr. Watermeier. I believe they are ready. They are working to improve it and to be ready for the expansion that is going to happen. But to talk about the resiliency specifically, in Nebraska we have had an event last year with the wildfires, and I spoke with several telecoms. They are basically unaffected because they had vaults, they had pieces of puzzle that were covered. When we had the floods and the ice storms in 2020, we were covered there, we lost very little infrastructure as far as that goes. So I have a lot of confidence that the telecoms, specifically in Nebraska, and across the country, are actually ready for the AI. And they are just anxious to do it, but we have got to be--worry about this last mile, and we have got to be worried about the last customer. Senator Fischer. Do you think these customers are going to have that connectivity, that they are going to need to compete in sectors like manufacturing and agriculture? We have heard it takes that fiber. What are we looking at here when you have the basics involved with AI and how that is going to be available? Mr. Watermeier. You know, infrastructure isn't the same in every city as it compares to the rural areas. In your home, I know in the counties where you live, we have the infrastructure there for manufacturing, which will be much easier. We have the infrastructure there for the first step in agriculture, someone's office, their home. But when we are going to talk about going over the hill with an autonomous vehicle, or when we are going to talk about going over the hill to look at something specific, on my farm I specifically use satellite and drone imagery to look at weeds, to look at different things. It tells me quicker than I could ever be there. We have that technology, but the next step is what I worry about. And I think the telecoms are ready, but I just don't think we are going to have the connectivity of the last mile, the last animal, the last acre like we would like to have. Senator Fischer. Can you tell us how important the Universal Service Fund and Nebraska's Universal Service Fund are to, ensuring that rural communities have that connectivity so that they can participate in the AI economy today and into the future? Mr. Watermeier. It is critical. I mean, the Nebraska Universal Service Fund is what we consider partnered with the Federal side. We watch the dollars that they receive federally, and the programs they receive. We try to match and fill those gaps. Nebraska has actually reached--and my philosophy is on everything I have done in the Legislature or in the Commission is all about infrastructure. I would like to spend a dollar, but not just have it help or serve one individual, but see how it helps the overall infrastructure of the country, the county, the city, the entire world really. Because that is the key for me when I think about building this infrastructure, is it is not just going to help one individual, but if we can do that, we will actually helped infrastructure for the whole country. Senator Fischer. We all know infrastructure is an investment in the future of our communities and our country as well. Mr. Spalter, many of your members utilize Universal Service Fund as well, so I put that question to you. How important is the Universal Service Fund to be able to have that connectivity to rural communities that need to, not just want to, but that need to participate in an AI-centric economy? Mr. Spalter. And I would add they must participate in an AI-centric economy, and one of the accelerators of that is a reformed, thoughtful USF program that can keep pace with the changes that we are seeing in our network, and the needs of the communities that it is designed to serve, which is why we are very grateful for your service as Chair of the USF Working Group, along with Senator Lujan. And we are going to do everything we can to support your work to ensure that every community can get the fiber infrastructure they need, maintain it, optimize it for the long haul so that we can actually deliver AI's benefits to every community in the country. Senator Fischer. Thank you. Mr. Everson, when somebody picks up their phone, they want to make a call or send a text, the person is usually not thinking about the nature of the network over which that data is being transmitted unless the call drops or unless the text doesn't go through, or the e-mail gets stuck along the way. This is largely true for end users of AI as well. Residential and businesses, they care about the quality of the data and the outcome that they have. However, they likely don't care about the network that allowed them to be able to use that AI model. So, Mr. Everson, can you describe the need for complementary network technologies that are going to support an AI economy, and especially the end users for that AI? Mr. Everson. Absolutely. Thank you, Senator. It is a great question. The way I like to think about it is exactly as you said it. The user doesn't want to have to care what network they are connected to. They want access to that service regardless of where they are. So whether they are on their Wi- Fi network at home, or they are on Wi-Fi at their company, whether they are out on the macro network, and they are using the cellular network, or they are using satellite services as were mentioned earlier, or they are at home on their broadband connection, it should all be the same. And that fabric is really essential to our success. Senator Fischer. Thank you. Senator Klobuchar, you are recognized for your questions. STATEMENT OF HON. AMY KLOBUCHAR, U.S. SENATOR FROM MINNESOTA Senator Klobuchar. Well, thank you very much, Senator Fischer. And thank you to all of you for being here today. I believe that our laws need to keep up with the technology and be as sophisticated as the technology and have a coordinated approach. I have been frustrated that we haven't been able to move forward on different fronts. Senator Thune and I hope to have our bill completed over the August time period so we can have a markup immediately when we come back to set some standards. And I know there are other ideas out there as well. Because we just can't stand down and not have coordinated rules in place when it comes to AI and when it comes to content with AI. I think this technology could bring us miraculous cures. I know Mayo is using it now in some really good ways in Minnesota. But as David Brooks once wrote, he doesn't know if-- how to write about AI because he doesn't know if it will bring us to heaven or hell. So if we do nothing, it is going to bring us to hell. And I think we could be in much better shape if we put some rules in place. So we are going to start with some of that with this committee. But I just don't think we can keep waiting until the next election, and the next election, and the next election. So I am going to start out just asking you, Mr. Ramzanali, about what steps need to be taken within the Federal Government to strengthen our oversight of AI that presents high risk, including the risk of cyberattacks? We have had a number of them, as other states have had. We just had some water structures that was--water infrastructure that were attacked. I don't know if it is AI related, by the way, cyberattack. Mr. Ramzanali. I appreciate your question and your long leadership on so many of the issues that AI governance includes. To answer the narrow part of it, which is how do we think about cyber and those kinds of risks, to me, there is two parts of it. You have to harden the infrastructure that critical infrastructure operators have. And you have to require pre-deployment testing so that the government and critical infrastructure providers know the scope of the threat and can mitigate risks. Those two parts both have to happen. Senator Klobuchar. OK. And Mr. Everson, in order to protect the fast pace of cyberattacks, we likely need to use AI to respond just as fast. So what steps should companies take to strengthen their own cybersecurity practices, including by using AI? Mr. Everson. Absolutely. Thank you, Senator. You know, there is no surprise that AI has accelerated the discovery of vulnerabilities and the access to those. You know, what we are seeing is also an ability using these models. If we look at things like Project Glasswing and Daybreak, we were actually able to identify, with early access there, our team gave me a fascinating stat, they were able to go through 1.8 billion lines of code in 5 weeks' time, which is unprecedented. So there is a balance that it gives the good guys really good tools to use. From our standpoint, there is some really basic recommendations to get to the point of your question. Number one is patching, patch, patch, you know, install the patches. Number two, is identify the products that aren't patchable. There is a lot of tech deck out there that needs to be upgraded. Multifactor authentication to limit people having access to the network. And then a zero-trust architecture which, you know, doesn't just rely on a perimeter, but everything that happens in there requires authentication. Senator Klobuchar. OK. Mr. Spalter, can you talk about the changes and updates our telecom infrastructure needs to support what is coming at us with AI, or what is already here? Mr. Spalter. I think that the fundamental physics of the technologies that comprise fiber technology are already in place, which is the symmetrical capabilities at scale, and at speed, and at low latency to be able to do the throughput that AI is going to be required, not only to support the infrastructure, but more importantly, human beings that are going to be needing it, which is why we need to deploy more fiber as more broadly as possible. And that is why I will go back, hopefully not as a broken record, to the essential need for permitting reform. I would also say that---- Senator Klobuchar. And just could you explain, you see this at the state level, Federal level, how do you see this going if you could wave a magic wand? Mr. Spalter. If I could wave a magic wand, we would move aggressively right now to pass the bills before Congress that you are evaluating, not only for Federal land, where we can at least have shot clocks and deem granted for previously analyzed projects. But also extend at the state and local level more efforts, some of which are being analyzed right now by the FCC, to apply to wired broadband, therefore fiber broadband, the same types of approaches on state and local permitting that we have been successfully able to do in 2018 on wireless networks, which is one of the reasons that we were able to scale 5G as quickly, and take global leadership on it as quickly as possible. Senator Klobuchar. Thanks. Just one last question, since it is for the witness from Nebraska, Madam Chair, if you don't mind? Mr. Watermeier, I know we have a mutual friend in Minnesota, and you know, could you just expand on what Mr. Spalter was talking about at the local level of what you think we should be doing to protect? As you know, there is some very small towns, and small counties, and small communications networks. Mr. Watermeier. You are exactly right, Senator. And thank you for the question. We have more than just one friend in common, I am pretty sure, in Minnesota. Mr. Spalter had reached a little bit of a point there, he talked about how the state's role is. And I don't want to get into the weeds in suggesting anything at the Federal level, but I do have a lot of confidence in the stakeholders that are here today, and many of them that aren't, that really know what they need to do in the weeds. But the states are going to have a role in this, I believe, and they should have. I mean, even our Commission deals with all the utilities, and we are talking about this very thing all the time with them. Do they have--generate enough revenue to cover their costs in cyber, and all the different things? And if a black sky event would happen, are they ready and are we talking? So I think there is a role in the states to do this. And I hope we are not preempted completely at the Federal level. So that would be my answer, Senator. Senator Klobuchar. Thank you. Senator Fischer. Thank you, Senator Klobuchar. Senator Klobuchar has always been a strong leader and advocate for many of these issues, and always for rural areas in the states, so thank you. Senator Capito, you are recognized. STATEMENT OF HON. SHELLEY MOORE CAPITO, U.S. SENATOR FROM WEST VIRGINIA Senator Capito. Thank you, Senator Fischer. And thank all of you all for being here. Mr. Spalter, you are singing my song here on permitting. I noticed in your statement, you said the primary constraint on America's AI infrastructure is not technological, it is governmental. Permitting is the single largest roadblock. Most people think of permitting reform in terms of energy development, or possibly transportation systems, but the communication systems and the AI systems are as important, and maybe in some cases, as you are making the case today, more important, to be able to move quickly and swiftly. You probably know, I am the Chair of the Environment and Public Works Committee. We have been in active negotiations with the Energy and Natural Resources Committee at the Committee Chair and Ranking Member level for really months now, to try to get to some of the permitting issues that you talked about. You mentioned shot clocks public lands. Are there other things that you think permitting reform would help in the industry to be able to build out as quickly, the infrastructure that is needed? Mr. Spalter. Yes, Senator. I think that there is several initiatives that we can actually move on very, very quickly. With respect to Federal lands, if we can simply allow previously analyzed projects to be deemed granted so that we can move swiftly to transit lands that are managed by the Federal Government. I think there is also going to be opportunities for us to through, not only legislation but also through regulation, as I mentioned before, to ensure that state and local communities can have the tools and leverage they require to be able to speed up permitting for wired broadband infrastructure by applying the same approach that we have successfully been able to do on the wireless side through FCC regulation. But we can't stop there. We also need to make sure that while we are actually speeding up permitting, we are not taking the foot off the gas and supporting the networks that we are finally able to deliver through a thoughtfully reformed USF mechanism. We are so grateful for your work in that regard. Senator Capito. Well, thank you. As you know, the BEAD funds are getting ready to come out. Some of them, I think, are out. It is taken way too long. We complain constantly about that in this committee and at home. But there is a thing called non-deployment funds, which are the funds that are not going to come out immediately because of the way this administration has reworked the program. So I would go to my own home state of West Virginia that has a lot of physical challenges in terms of being able to build out more expansive, and so the actual deployment and permitting is more difficult for us in some ways. Senator Wicker has had a bill that I am on, on the non-deployment funds, to allow those to come back to the states, the original funds we were supposed to get, to come back to each individual state, so we could use this for the leftover BEAD money that would strengthen these AI networks and let that be a permissible use. Do you have an opinion on that, Mr. Spalter? Mr. Spalter. Yes, we do. First of all, we all appreciate the work that you and Senator Wicker did to advance that bill. We think that there is a need for speed to delivering guidance to the states on non-deployment funds and their uses, I know the states are waiting, and we have heard from our colleagues at NTIA that that guidance will be forthcoming, we hope, in September. Senator Capito. Right. Mr. Spalter. We believe that any dollar should be accelerators of the fundamental purpose of the BEAD program and we supported this body's determination that fiber should be prioritized in that program. Nevertheless, we think that those dollars can be used purposefully for things like resiliency, cyber training, focusing on using dollars for accelerating permitting, and also for modernizing our NG911 architecture, all of which will be additive, and turbocharge the ability to optimize fiber deployments. And also every other technology that will ride on them, including satellite. Senator Capito. Yes. I do think it fits quite nicely with the permitting reform effort as we see it moving. Mr. Everson, do you have an opinion on that? Mr. Everson. On permitting? Senator Capito. Well, permitting I would love to hear, because I am sure you all are involved with that. So let us start with permitting, and then non-deployment of funds. Mr. Spalter. Oh. OK. Thank you, Senator. So on permitting, you know, I kind of look at this from maybe a simplistic perspective as an engineer. There are a lot of things that need to be done to build out the infrastructure that is required and so the sooner, the more that we can speed that permitting process along the better, because then we can get to the work of deploying those resources. From a BEAD standpoint, what I would say is that, you know, number one, we need to make sure that everyone is connected and everyone can share---- Senator Capito. Right. Mr. Spalter.--in the good fortunes here, and so we want to see that. And then if there is any funds left, we would love to see that invested in the resiliency and security of these AI networks. Senator Capito. All right. Thank you all very much. And I look forward to, hopefully, having that permitting reform bill across the line. Thank you. Senator Fischer. Thank you, Senator Capito. Senator Rosen, you are recognized. STATEMENT OF HON. JACKY ROSEN, U.S. SENATOR FROM NEVADA Senator Rosen. Well, thank you, Chairman Fischer. And I want to thank you and Senator Lujan for holding this hearing. Witnesses, thank you for being here today. I want to talk a little bit about AI networks and fiber, because we have heard from witnesses today how important high speed, low latency connectivity is for AI networks. So Mr. Spalter and then Mr. Ramzanali, what are the biggest barriers you see to our network infrastructure meeting the moment on AI? Mr. Spalter. I can begin. Well, thank you for that question, Senator. It is good to see you again. I think the primary barrier that we have to accelerating the AI opportunity, not in terms of only its scale and development at the infrastructure level, but as importantly, its ability to actually have a positive impact on human beings, in Nevada and beyond, is the perennial challenge that we are still facing on sluggish permitting. The question isn't whether we get the approvals. It is the delay in the time that it takes. And there are many levers that you have, Senator, to actually pragmatically and thoughtfully accelerate the ability for us to get the permits that we need to build fiber broadband into your rural, and other communities. Senator Rosen. Thank you. Mr. Ramzanali? Mr. Ramzanali. I will agree with that and add that when I look at it from the highest-level perspective of the network, the failures are that it doesn't reach all of the people in Nevada and around the country. It needs to reach rural communities. It needs to reach low-income communities. We need to make sure that the highest communications networks are available to everyone. Senator Rosen. Thank you. I appreciate that shout out for Nevada because we are the seventh largest state and the most mountainous state in the Lower 48. So getting out to those deep rural communities is particularly important. And I wanted to say I agree with my colleagues because we hear about the issues with permitting in Nevada. More than 87 percent of our state is Federal land, 87 percent. And under this administration, we have unfortunately seen unprecedented cuts in staff to the Federal agencies that approve the permits for telecom networks across Federal land. Agencies are understaffed, under-resourced, many don't have the technical resources to act more efficiently. So Mr. Spalter, with this environment in mind, will you continue to work with Congress to support Federal agencies and make the permitting process more efficient for projects with limited, if any, environmental impact like any telecom projects? Mr. Spalter. Absolutely. The number of horror stories that we have heard about Federal land permitting are incredible. It is almost like you can sit around a campfire with a flashlight under your face. Senator Rosen. Yes. And 87 percent of public land, it is really about sluggish. It just stops us. Mr. Spalter. But the good news, though, is that you have very solid legislation to advance more thoughtful, pragmatic, common-sense permitting. And that there are efforts, and we are talking to the agencies about extending things like categorical exclusions for previously analyzed projects. There is more that we can do, but there is a fierce urgency of now and getting it done. Senator Rosen. Thank you. And I am going to talk a little bit, too, now about network cybersecurity. Because network intrusions, we know they are increasing. And of course, with increasing AI sophistication and use by, well, everyone, our networks are going to be more and more vulnerable to sophisticated cyberattacks. So what I really want to ask is that, to Mr. Ramzanali, how do we avoid a comprehensive and undetected cyberattack across our networks, like what we saw with Salt Typhoon? Mr. Ramzanali. I think that is a great question. Salt Typhoon was pre-AI. It was low-tech. They used stolen passwords and equipment that wasn't updated. The way to think about this is, as I said earlier, the two sides of the coin. We have to harden our infrastructure. The cyber defenses have to go up. And that means enabling critical infrastructure operators to use AI to their defense. But it also means pre-deployment testing on AI models so that we know what the threats that are coming look like. Senator Rosen. Thank you. And would you agree that we should really be staffing up and strengthening CISA, our Cybersecurity Infrastructure Security Agency, that protects-- helps to protect that public-private critical infrastructure? Mr. Ramzanali. I completely agree. Without a fully staffed government, that has agencies like CISA, but also the Cyber Safety Review Board, there is many parts of government that need to be strengthened and given the capacity to deal with today's moments. Senator Rosen. Thank you. I appreciate that. And I just want to ask one last question to Mr. Spalter. What can Congress do to help ensure the lack of network capacity does not widen the digital divide as AI tools become more integrated into everyday life? We are seeing it every day. What suggestions do you have for us? Mr. Spalter. I have three. One, is to ensure that we can move forward on thoughtful, swift permitting reform. Second, is to use the USF Working Group process to ensure that the future networks that are going to be required to meet the moment that AI capacity is going to require, will be thoughtfully worked through. And the third, is another piece of CISA, is to ensure the security of our networks, in so doing by making sure that you can reauthorize CISA 2015 so that we can talk to each other, and to government, to ensure that the networks that we are going to be deploying to the communities that need them can be as secure as possible. Senator Rosen. Thank you. I appreciate that. I yield back. Senator Fischer. Thank you, Senator Rosen. We have been joined by the Chair of the Committee. Senator Cruz, you are recognized. STATEMENT OF HON. TED CRUZ, U.S. SENATOR FROM TEXAS Senator Cruz. Thank you, Madam Chair. We are in a strategic competition with China to lead the world on AI. And winning that race requires massive capital investment to build the networks that will power tomorrow's AI applications. When we talk about permitting, it is often on the energy or resources consumption fronts. But building out the networks to connect new AI infrastructure faces many of the same hurdles. Mr. Spalter, first of all, happy birthday. I understand today is your birthday. Mr. Spalter. Thank you very much, Senator Cruz. Senator Cruz. So we held a giant hearing for you in celebration. For your members, to what extent are outdated, slow, or duplicative permitting requirements at the Federal, state, and local levels slowing broadband network deployment? And how many years, on average, does fiber construction take to reach a given rural or remote location? And what percentage of those delays are due specifically to permitting? Mr. Spalter. It is the singular biggest barrier, block, and pebble in the shoe of our innovation economy. Fiber networks can only be built at scale and at speed if we can equally give attention to permitting reform. We can deploy networks in weeks, sometimes in days, we are ready, equipped, and have the capital and the will to do so. The challenge, though, is that we are seeing, in many instances, that we are having to sit and wait for months, sometimes years. In Big Bend, one of our members had to wait 2 years to get a permit from one of our Federal land agencies over a debate about whether there was a forest in the Chihuahua Desert. This is illustrative of the kinds of challenges that we are facing at scale, which is why the legislation that you are considering here is so vitally important. Senator Cruz. Well, thank you. And I can verify firsthand, there is no forest there. Mr. Spalter, in addition to new builds, your members are investing heavily in upgrading their existing networks to keep pace with growing data traffic. Are permitting delays slowing those upgrade projects as much as new construction? Or are these bottlenecks different from what you see with greenfield deployment? Mr. Spalter. Permitting delays are an equal opportunity offender. We are seeing both for new projects but also for upgrades, where simple projects that will require simply replacing, literally, a strand of fiber can sometimes be delayed for months. We have seen that in states like Utah, where we could easily establish new approaches, either categorical exemptions for broadband, stepping forward to allow previously analyzed projects that have been previously permitted to proceed on a deemed-granted basis. These are simple, practical, progressive solutions that actually can allow us to move forward as a nation, and not allow the Chinese to get out in front. Senator Cruz. Well, I hope your words prove prescient. And I hope that they foreshadow bipartisan agreement to get this done. Let us shift. Mr. Everson, beyond network efficiency and resilience, how are Cisco's AI-driven networking technologies helping your carrier customers differentiate their service offerings in a competitive marketplace? Put differently, are you seeing AI create new service offerings rather than just providing faster or more resilient networks? Mr. Everson. Thank you, Senator. And I am so glad you asked this question, because this is one of the exciting things about this. As AI shifts from primarily training dominant to more inference dominant, there is a huge opportunity for telcos to play an important role besides just carrying the bits. And so we see them providing services in the tokens' generation side of things. One of the exciting things we are doing with AT&T, in our home state initially, is we built out a service offering called AI Grid, where they are actually serving AI capabilities, AI models, and AI outcomes from their network to their enterprise customers. So it provides low latency, very high secure services, and allows them to move up the stack to deliver value. Senator Cruz. And what does this mean for economic growth and for jobs? Mr. Everson. It is actually a really fantastic story for that. Because now they are able to move up that value chain, deliver more value to their end customers. It makes the employees more efficient, more effective with these AI models enabled there, and will drive growth. Senator Cruz. AI workloads and use cases range widely, from optimizing networks, to powering machinery on factory floors, to enabling real-time diagnostics at rural hospitals, each demands something different from the network. There is no question that fiber is a wonderful technology under ideal circumstances. But given how long buildouts would take at a nationwide scale, putting all of our eggs in that basket would leave many innovative network use cases, not to mention entire regions of Americans relying on them, waiting for years until shovels broke ground. Mr. Everson, can you speak about how embracing the full toolkit of network technologies available today, including fiber, but also wireless, satellite, and others, will enable greater connectivity and access to AI across all America, not just in data center hubs, and on a faster timeline than a fiber-only approach? Mr. Everson. Thank you. And I couldn't agree more. You know, I mentioned earlier, but the user shouldn't have to care where they are connected or how they are connected, they should get the same level of service. That requires this fabric of cellular connectivity, Wi-Fi connectivity, fiber connectivity, satellite, all of these need to work the same. Senator Cruz. OK, final question. Mr. Everson, Cisco has visibility in the network traffic patterns across the industry. Based on what you are seeing, is the 800-megahertz spectrum pipeline, and the Working Families Tax Cut sufficient to meet the U.S. carrier's 6G capacity needs? Mr. Everson. So number one, thank you to everyone for the 800 megahertz. That is enormous, and a global imperative, but obviously a domestic imperative. I think what I would say is, it is a great start and a significant chunk, and merits further evaluation after that. Senator Cruz. Very good. Thank you. Senator Fischer. Thank you, Senator Cruz. Senator Blunt Rochester, you are recognized. STATEMENT OF HON. LISA BLUNT ROCHESTER, U.S. SENATOR FROM DELAWARE Senator Blunt Rochester. Thank you, Madam Chair. And thank you also, Ranking Member Lujan. And thank you to the panelists. Mr. Everson, AI applications are becoming more integrated into everyday life, from education and health care, to businesses and manufacturing, and as they do, Americans will increasingly depend on broadband networks capable of handling the demand. Cisco has projected that AI and Agentic AI will drive consumer network traffic up by roughly 6.6 times by 2035, about 64 percent more growth than we would see without AI. What investments or network upgrades will be necessary to ensure broadband networks can keep pace with the AI-driven demand? Mr. Everson. Thank you so much, Senator. The capacity is one story, and you illustrated that very well. And so we need to keep upgrading and enhancing our fiber capacity, cellular network capacity, all of these different pieces. We also need to account for sort of the differing traffic patterns as well. So while we are reengineering those, you know, the traffic pattern is shifting to where, you know, primarily we are consuming content on our phone or, you know, at home or whatever. Now, as that context and everything has to go up to the cloud, we need more lanes going upstream as well. Senator Blunt Rochester. Thank you. Mr. Spalter, you have written that: Only fiber offers the superior capacity, reliability, and future-proof scalability that AI demands. I would love for you to expand a little bit more on that for the record, and talk about what policymakers should keep in mind as we are looking at the long-term tradeoffs between fiber and other broadband technologies, as we increase AI-driven economy? Mr. Spalter. Thank you for that question, Senator. AI- related traffic will increase demands across every layer of our communications ecosystem, not just fiber, but also Wi-Fi, satellite, mobile, and beyond. What is essential, though, as we think about powering the pump on allowing greater capacity across our communications ecosystem. is that it will always be fiber that will become the backbone for every other technology to transit their higher capacity and lower latency traffic that AI will require. So in policy terms, this means that while we have to create incentives for evolution of all technologies, we must do more through things like USF reform permitting to get more fiber as extensively as possible so that it can support every technology that has applicable uses. Senator Blunt Rochester. Thank you. And, Mr. Ramzanali, the Trump administration's rewrite of the BEAD program's rules didn't just open the door to satellite as a lower-cost alternative to fiber, but it also deemed some homes and businesses already connected, excluding them from funding eligibility altogether. President Biden's NTIA allotted enough BEAD funding to reach nearly 6,000 homes with higher-speed Internet and fully connect Delaware. Our state would have been fully connected. If I could ask you, the Trump administration has cut that funding by about 74 percent. This disproportionately impacts rural communities like those in Kent, and Sussex, County and Delaware. Looking ahead, how will lack of access to reliable high-speed Internet impact Delawareans' ability to participate in an increasingly AI-driven economy? Mr. Ramzanali. Thank you for that question. If we don't connect those thousands of homes in Delaware that are not connected, it means that they are missing out on the economic benefits of not just AI but access to the internet. It means that when their kids have homework assignments that they need to access the Internet or to use AI, they don't have that access. It means when telehealth visits, or being able to ask questions of various tools, they don't have that access either. So these are real-world limitations that are being put on thousands of people in Delaware, and around the country. We shouldn't accept that. We shouldn't accept the state of the country where not every American is connected to high-quality networks. Senator Blunt Rochester. Thank you. I can tell you when folks make decisions about what is eligible and what is not, if you are just driving through one of our counties, you can already predict where your call is going to drop, where you are not going to have access to service. And to your point, this is the cost of doing business. This is the cost of our health care. This is the cost of even connecting to our neighbors. For me, making sure that we are all connected is a priority. And I thank you, I thank the panel, and I yield back. Senator Fischer. Thank you, Senator. Senator Young, you are recognized. STATEMENT OF HON. TODD YOUNG, U.S. SENATOR FROM INDIANA Senator Young. Mr. Spalter, and others, welcome. Mr. Spalter, we are going to cover a lot of ground here, so prepare, please. Indiana just landed a quantum tech hub under the CHIPS and Science Act, and our state is poised to play a critical role within the Nation's quantum corridor, as it is called. With significant investments in both quantum computing and artificial intelligence, how important is robust fiber infrastructure to support, not only AI, but also next- generation quantum technologies? And what does that mean for America's global competitiveness? Mr. Spalter. It is not just indispensable, it is existentially important, to not only the success of the quantum corridor, but to our ability to scale next-generation quantum networks in such a way with speed that we can actually outpace our very well-funded and determined competitors, including the Chinese. We just, yesterday, had a webinar with the Director of the White House Quantum Coordination Office, and we explored exactly what are going to be the policy levers that can accelerate our ability as a nation to win, including in Indiana, and that is going to be about prioritizing more fiber buildout and policy mechanisms that can incentivize it, including permitting reform. And so we are really eager to work with you on passing various bills before this committee to get that done. Senator Young. I am so glad you checked permitting reform there as well. I will say, and I advertise this to all that will listen, I think we are very close in terms of bipartisan coalition. I think it needs a nudge from the powers that be to get it across the line, maybe make some principled accommodations, but I think we are close to enabling that golden age the President talks about. Long-haul fiber networks use facilities called fiber huts, to keep signals strong over long distances. These important facilities are often left out of our conversations related to communications infrastructure. Because fiber huts are sited in many states and local communities, different rules for permits and construction can slow down new projects and broadband deployment. How do different state and local rules affect the building of these facilities? And most importantly, what can Congress do to make the process faster and more consistent? Mr. Spalter. We are going to see increasing innovation as to how we can accelerate fiber optic networks and optimize them, including fiber huts. Oftentimes, these can be as small as eight by six feet placed in a rural area, yet many times rural--state and local rules will treat that tiny fiber hut in the same way that they might have a significant large advanced manufacturing facility in terms of permitting processes. If we can move forward swiftly to pass legislation like CLOSE THE GAP Act, the RAIL Act, Accelerating Broadband Permits Act. And ask every committee of Congress that is considering permitting reform to include broadband in its consideration, we will be able to actually get the work done. Senator Young. Fantastic. Two groups of questions, 2 minutes left. Here we go. Cybersecurity and the risk, perhaps increasing risk, that bad actors with access to sophisticated advanced frontier AI models will utilize them for nefarious purposes, including to initiate cyber or ransom attacks. How important, sir, is the ability for government and the private sector, including broadband providers, to share cyber threat information in real time? Mr. Spalter. The clock is ticking, sir. We are currently authorized through CISA 2015 to do that information sharing through September 30th. It needs to be reauthorized for the long haul, given the increasing sophisticated and advanced threats that we are facing. Without that ability embedded in CISA 2015, we will have not just one hand tied behind our backs, but two hands tied behind our backs. Senator Young. Well, you anticipated my last question, talking about the expiration of CISA 2015. So I will give you in the remaining 45 seconds an opportunity to unpack a little bit more, if you can, how expiration would impact potential mitigation and response capabilities. Mr. Spalter. How, I am sorry, what? Senator Young. How expiration of CISA 2015---- Mr. Spalter. Yes. Senator Young.--would impact mitigation or response capabilities. Mr. Everson seems eager to offer a response. Is that accurate? Or were you just adjusting yourself in the chair? Mr. Everson. Thank you, Senator. You know, one thing that I will say, I am probably not the best person on the policy, but the thing that I will say is that your point around threat intelligence is one of the number one priorities that we have. At Cisco, we actually--one of our big strengths is that we have an enormous threat intelligence organization. If we didn't have that, we would not have the level of sophistication on cybersecurity. Mr. Spalter. If I could also, that one of the reasons that we need to keep a pace with AI innovation and not slow it down, is because it is the very AI tools and models, both frontier and open weight, that are becoming part of our defense mechanisms. We need to be able to advance them in order to be able to protect our networks against very well-equipped, very well-financed, resolved adversaries. Senator Young. Proving, once again, these are national security issues. I keep saying this committee is a national security committee, so when I am working on merchant shipping, Department of Commerce, it is a national security issue. It is a national security issue. This is not about creating jobs. It is a happy byproduct of a national security play. Thank you, Chair. Senator Fischer. Thank you, Senator Young. We have been joined by the Ranking Member of the Committee, Senator Cantwell. Senator, you are recognized. STATEMENT OF HON. MARIA CANTWELL, U.S. SENATOR FROM WASHINGTON Senator Cantwell. Thank you, Madam Chair. Thank you so much to you and Ranking Member Lujan for holding this important hearing and your continued focus on these important communication issues. I want to thank my colleague from Indiana who just spoke about how this is a national security issue, and he and I worked on the first Federal role for our government in AI policy and got that implemented, and every day I am thankful that we got that started, I think, almost, I don't know, now I think it is probably five--six years ago. But anyway, we are here today, and we continue to see a fast-moving environment. And the question is what are we going to do to step up to that, both from the security point of view and from a competitive point of view? So artificial intelligence, we saw last week with OpenAI's disclosure of an unprecedented security incident. One of its advanced models broke out of the sandbox and created a complex cyberattack against Hugging Face. So AI can design viruses, toxins, biological weapons. It can be used to design an entirely new pathogen that are difficult to detect and counter. And today's hearing comes at an important moment as AI accelerates new capabilities that are being unlocked every day. But we still need to compete. So Mr. Ramzanali--is that how you pronounce it, [Ram-za-na- li?] You brought this up in your testimony about the Hugging Face making it abundantly clear that we needed to have regulation on these, not self-certification, but some regime, and that it might often need us to look into this from an international perspective, too, because we are not just regulating this technology for what might happen in the United States, but what maligned actors might do with this technology as well. So what do you think that we need to do to make sure that we get this right, particularly--I think this committee knows a lot about cybersecurity, I think over the time I have been on the Committee, we have had many, many hearings about cybersecurity. I wouldn't say we necessarily, it is a whole-of- government response with many different agencies, and we continue to see attacks on the private sector, telecom sector, ill-prepared to respond. And we also saw it on our pipeline system. But that makes that look like child's play--this makes that look like child's play. So could you elaborate on what you think we need to do to address issues like the biological threats, where AI could provide a step-by-step instruction for engineering of novel, resistant, supercharged superbugs? Like that could just go off and do that. The advanced AI models could be really key in solving some very important programs, but also could be used for uranium enrichment and nuclear bomb trigger designs. And so if that gives a lot of people, on an international basis, access to more information and proliferation, what do we do about that? And then just on this cybersecurity model that, again, very familiar coming from a very big software state. We all know that software can be buggy, but now we are not talking about buggy software, we are talking about software that writes itself, goes attack systems, creates big holes. It is almost like it is a chemical itself in the kind of damage that it could be doing corrosively within our cybersecurity network for safety, and we wouldn't even know about it. So that is a lot, but we invited you here because we thought you could help us with this. And what are the things that we need to put in place? Mr. Ramzanali. It is a lot, but this is the core of the problem. So I appreciate you highlighting it and your longstanding work on these issues. I think about this in a few different ways. First, all of the critical infrastructure systems that we have today need to be hardened. Second, pre- deployment testing can help us become aware of the types of threats that are coming. And this is the role of government. Government has biologists. It has nuclear weapons experts. The national labs that we employ have those kinds of experts. We should be deploying all of them to help understand what the threat landscape looks like. Third, each one of those threats that you described has its own supply chain and ecosystem. And we need to look at it from a systems lens. When we talk about the biological concerns that you raised, there is a question about what are models able to do? What does that look like when you are able to do one exploit after another? But there is also a whole host of other things that have to happen, including integrating with cloud labs, getting access to biological agents. So we should be looking at every part of that and worry--I am worried that we are not looking at every part of that and doing as much as we can. Both the last administration and, to its credit, this administration have made progress on some of these questions. But we have to do way more. I think you are exactly right to say that the landscape is changing, and we are still moving incrementally in some of these domains. Senator Cantwell. We shouldn't be fearful about these national security organizations within our government being involved in trying to get this right and protect us. Is that correct? I think some people are proposing there is like this self-certification model, and we just have to go fast. Mr. Ramzanali. I think this---- Senator Cantwell. And I think what we have to go fast on is all of us coming together on what are the right security layers to protect our citizens, and our nation, and to advance our competitiveness. And so that often does take some of the smartest people within our Federal Government, particularly when you are dealing with some of these biological or nuclear weapons. Mr. Ramzanali. That is exactly right. The scale of the problem we are talking about when it relates to safety and security is not commensurate with a self-certification voluntary regime. We don't put up with massive safety national security concerns with paperwork being the response in any other domain. Senator Cantwell. Well, it certainly hasn't worked out in aviation, I can tell you that. But we are just advancing too fast, if you ask me. Even in aviation, you have to have--if you are trying to understand the interaction between hardware and software. But in this case, I think that what hasn't been amplified is how dangerous the cyberattack universe is, just that in and of itself. And how promulgating, I guess is the right word--I am not sure that is the right word, but I am going to use it, promulgating of the attack that AI could be. It could be quite significant on its own, you know, going through a system. Mr. Ramzanali. It would be incredibly disruptive to our lives if we have to consistently worry about AI-based cyberattacks disrupting banks and hospitals and telecommunications systems. Because that is just happening as a regular part of our life. We should not accept that future. We should try to avoid a future where those kind of attacks become commonplace. Senator Cantwell. Well, I don't think we have quite got the cyber regime correct yet today, because we are still seeing attacks, and a lot of personal information still being exposed. And I think we continue to want to work with our telecom industry to make sure they are making the investments that keep us going. But just I see I am over time, Madam Chair, can I go on one more? Senator Fischer. Um-hum. Senator Cantwell. How do you suggest that we--what areas do you suggest we engage on an international basis to get an international regime? What areas would you suggest? Mr. Ramzanali. The international part is the hardest. That is where there is going to be diverging interests that we have to figure out. One of the oldest existing international organizations we have is the ITU, which started as the International Telegraph Union, because we needed to come together to figure out a system that works for a technology that doesn't respect borders. That is where we are today. And the threats are much bigger, and the issues at play are much larger. We do have to figure something out. There was a lot of work done in the last administration. This administration is doing a lot. All of those feel insufficient. So I think we need to rethink how we approach this and start from what are everybody's interests. And I do believe there are shared interests in safety and security. Senator Cantwell. I agree, but I think we have to think about some of these nonproliferation regimes that existed, and how they functioned, and what we have to do here as well. And so we will look forward to more questions. But thank you for this very important hearing. Appreciate it. Senator Fischer. Thank you, Senator Cantwell. Senator Lujan and I are going to ask just a couple more questions before we close the hearing. Artificial intelligence has really presented a new array of challenges and opportunities for networks, and one challenge is the change in the profile of network traffic that AI has prompted. AI is not just a new source of network demand. It is also a new type of network demand that is diffusing throughout our networks here in this country. So Mr. Everson, could you describe how and why AI is not just adding demand to networks, but it is also changing the traffic profile, the traffic patterns networks are seeing? Mr. Everson. Yes. Thank you, Senator Fischer. And it is a great question. I would love to expound upon it. With AI, you know, traditional networks were designed really for kind of serving content downstream largely. There is some symmetrical nature to it. But to a large extent, the predominant traffic is downstream. When you are using AI, part of the value of AI is what you feed the model to give you a good answer back. So it is called context. It is whether it is videos, it is photos, sensor data, maybe just a bunch of text that you need to feed up in there. And it kind of goes back and forth. In an agentic world, that agent may go back and forth multiple different times to seek from different sources. And so it is changing the traffic pattern with more upstream, because you have to send more up to the model wherever it is, whether it is at the edge of the network or whether it is in a central location. And it is also changing the persistence, and the length of time of these transactions, and the traffic pattern substantially. Senator Fischer. How are networks adapting to accommodate those changes? When you look at the different patterns, that new type of traffic that is out there, how is it working? Is it working well? Are you having issues with it? Where is it working, and where are you pretty ill-equipped, I think, maybe at this point in time, too, to handle all that? Mr. Everson. So the good news is that we are still in the early days of inference. And so the impact has not been as extreme. Now, the bad news is that, obviously, it is growing very, very quickly there. You know, there are different mechanisms. And so we have talked a lot about bandwidth, and the need for fiber, and the need for spectrum, and licensed spectrum to allow for more uplinks, there are also mechanisms that the operators can put inside of their networks to help, sort of, offload or diminish that traffic requirement, where they can be involved in the inferencing chain. Serving tokens maybe with intelligent direction of requests to different models, and maybe by hosting some of that inside their network. So I guess I would say we have multiple tools in our toolkit that we are working through. And I think they are all going to be required. Senator Fischer. Thank you. And Commissioner Watermeier, what opportunities do you see for AI to improve emergency communications while making sure that our dispatchers remain at the center of the critical decisions that are going to be made when we have to have an emergency response? You talked about the wildfires in Nebraska, the bomb cyclone, the flooding, a lot of weather events that our emergency responders go through. So what do you see there? Mr. Watermeier. Well, thank you, Senator, for the question, because Nebraska does administer the underlying network of the 9-1-1. And we are really concerned, we are just at the edge of converting everything over to the next gen. AI is going to help in that in every single capacity. And even at the local level, it is going to help the training process. It is going to help the decisionmaking. And the tremendous amount of data that we are collecting when we switch to the next generation is going to be--AI is going to play a big part in that. Thank you for the question. Senator Fischer. Thank you. Senator Lujan. Senator Lujan. Thank you, Madam Chair. Mr. Everson, I want to pick up where Chair Fischer was asking some questions about traffic. As everyone has highlighted today, networks are built for traffic predominantly to come down, not to go up with AI. I appreciate your response to Chair Fischer about existing technologies. Those are Band- Aids. Traffic patterns are going to change dramatically. They already have. And in the United States, networks are still being built predominantly for download, with very little up. In New Mexico, I pay--I won't say the company--for certain speeds, and I don't get them, which is also the problem in most parts of America. I think there is a bunch of crooks out there that should be held accountable, and there should be penalty on companies that are not delivering the speeds that they are promising people. I have a bill for that. I have not introduced to it yet, but it is coming. So for all the companies that are in the room today and that are listening, if you are stealing from people-- I am trying to be better with my language--stop it. And if you know that your networks are going down and you are taking from people, refund them. Give them money back. That is not my question to you. What are the implications if the United States continues to build out networks that predominantly are built for download as opposed to the requirements that we are seeing with the uptake with AI and other information with the flow going up? Mr. Everson. Thank you, Senator. So plain and simple, we have to change, right. We need to get more symmetrical in the network architecture, in the underlying infrastructure, in the fiber access, in the broadband, in wireless. And so that plain and simplest part of it. There are tools as we get into AI where AI is going to make the network more efficient, we are developing agentic ops where the network can be more, you know, flexible and adaptable. But if it doesn't have that underlying infrastructure, that underlying backbone, then it doesn't work. Senator Lujan. And Mr. Spalter, my colleague from Indiana was asking some questions pertaining to quantum. We also know that as a result of faster speeds today, that that also increases the likelihood that deep fakes, people that are out there doing bad things, preying on our loved ones or ourselves, are able to do it at an increasing pace. Quantum networks are also promising more cybersecurity, quantum security as the United States builds out AI-driven network practices, protocols, we are looking at these other networks and policy, should quantum networking be part of the toolkit that the United States is requiring to be built alongside it as a future consideration--or sorry, not as a future consideration, but as something that we should be pushing to be implemented now? Mr. Spalter. The answer is absolutely, yes. I think that as we are envisioning the future network and the future needs that the network will support, including the types of traffic patterns but also use cases that quantum will deliver. We need to begin that work now. Not just in terms of the technology evolution that we are seeing in the labs and now transferring to commercial uses because other countries, particularly China, is advancing those efforts as well, but also in terms of our thoughtful ways of developing legislative and regulatory frameworks that will prime the pump in allowing us to deliver the promise of quantum so that on the long term we can do things like USF reform. We can do things like make sure that we are prioritizing all incentives possible including permitting reform to get more fiber, and its symmetrical speeds, and particularly importantly its low latency measured in the single digits of milliseconds that fiber networks will be able to deliver for our quantum future. Senator Lujan. I appreciate that. And Mr. Everson in one of your responses I think to the Chairman of the Committee, you said something along the lines of people--users may not be too concerned about how they are connected as long as they are connected to fast speeds. Should users be connected if they are connected? It is a loaded question. What I mean by that, Mr. Everson, is if people are not connected, I think people should care about that. There are people across America today that don't have access to the internet. Some because there is no connection, there is no offering, others because they can't afford it. And I appreciate what I believe I have heard from everyone about their support for USF reform. I would call it USF modernization. with all the work, with all the thoughtfulness that has gone into this space. My question to you, I have two closing questions here is, one, do you believe that USF modernization is needed? And that the United States should adopt it, and that Congress should act? Mr. Spalter? Mr. Spalter. Absolutely, yes. And we appreciate the work that you are doing, and you are doing Senator Fischer, in advancing that project. Senator Lujan. Commissioner. Mr. Watermeier. Certainly. And a good question Senator Lujan. I think about the network in the United States, it is like the interstate system built in 1956. 50,000 miles were built under cooperation. No private partnership necessarily. We need to be there. And I think USF really needs to be modernized, as you mentioned, and we are prime for it now. I think the communities and the citizens are ready, and they want to support it. Senator Lujan. Appreciate that. Mr. Everson. Mr. Everson. Yes. Thank you, Senator, and thank you for the clarification. Yes, I think USF has been an essential resource and something that we need to continue to look at and modernize for the AI era. Senator Lujan. Appreciate that. Mr. Ramzanali? Mr. Ramzanali. Yes, and we need to continue to focus on its effects on low-income consumers and how they can be supported. Senator Lujan. You know, I appreciate that very much. Look, as we get to this place, with the concerns that have been shared around the policy and what the United States needs to do to ensure that there is guard rails, and that there is rules, and there is guidance. That it is clear that there is liability whether it is to the companies that are developing these models that have responsibility with these agents, or it is the end user who is then taking an open model and then using it for nefarious reasons. There needs to be liability somewhere so that people don't do that, whether it is criminal, it is civil, or it is both. I also want to ensure that everyone across the country has the ability to take advantage of the promise of what this technology is. And that means we have to be able to get connected. And so I hope that we can preserve the investment, that we can see USF modernization, that everyone involved, including the FCC, does good rather than bad, that we move things in a positive way as opposed to rolling rules back that make it harder for everyday folks across the country. And Madam Chair, I want to commend you and the staff for this important hearing, for all the panelists that are here today. This is so important. I hope we have more conversations in this space, and that good policy comes out of these conversations. So just to applaud you for your efforts. Thank you. Senator Fischer. Thank you, Senator Lujan. As I said earlier, it is a joy to be able to work with you on these issues. You are very informed, you are very passionate, and we agree on many, many issues. So I am feeling positive on what the outcome is going to be. I would like to thank all of our witnesses for their testimony here today. Senators will have until the close of business on August 6 to submit questions for the record. And then the witnesses will have until the close of business on August 20 to respond to those questions. This concludes today's hearing. And the Committee stands adjourned. [Whereupon, at 11:37 a.m., the hearing was adjourned.] A P P E N D I X Response to Written Questions Submitted by Hon. Maria Cantwell to Jonathan Spalter Future-proof Networks. We need safeguards to protect consumers from AI's risks, but we also need to build the infrastructure that makes AI's benefits real for all Americans. Artificial intelligence depends on high-speed, low-latency connectivity, especially fiber to homes, businesses, and anchor institutions. Yet this Administration has spent eighteen months slowing broadband deployment to a crawl, steering funding away from fiber toward slower, less reliable connections, and sitting on guidance states need to prepare their networks and communities for AI. Earlier this month, NTIA Administrator Arielle Roth gave a speech titled ``There Is No Such Thing as Future-Proof Technology,'' arguing government shouldn't favor fiber because it could be leapfrogged. Question 1. Mr. Spalter, is there any technology on the horizon besides fiber, such as satellite, 5G, 6G, that can deliver the speed, capacity, and reliability that homes and businesses need for AI? Answer. No. While, every communications technology has a role to play in connecting Americans, when it comes to supporting our AI future, none replace fiber. They complement it. Wireless networks rely on fiber backhaul to connect cell towers, and even satellite traffic ultimately reaches terrestrial networks connected by fiber. As AI applications become more data-intensive, latency-sensitive, and increasingly dependent on high-capacity, symmetrical connectivity, fiber provides the scalable foundation for the entire communications ecosystem. Question 2. Do you think NTIA's current policies are helping or hurting America's ability to lead on AI? Answer. America's leadership in AI will depend on our ability to deploy the future-proofed fiber broadband infrastructure that enables AI as well as its reach to every community, business, school, hospital, and household. That means policies that accelerate investment, reduce unnecessary deployment barriers, and provide states and broadband providers with timely guidance and regulatory certainty. AI is fundamentally changing the demands placed on communications networks. Meeting those demands requires continued investment in high- capacity and ultra-low latency fiber infrastructure, modernized permitting processes, and efficient implementation of broadband programs. USTelecom's members have been working to deploy broadband across the country as quickly and efficiently as possible through massive private investment as well as participation in broadband funding programs. However, the biggest impediment to deployment is permitting. Delays and uncertainty in the deployment process slow construction, increase costs, and postpone the economic, educational, healthcare, and public safety benefits that modern broadband infrastructure makes possible. That is why USTelecom supports legislation such as the CLOSE THE GAP Act, and the RAIL Act, and the as well as the Accelerating Broadband Permits Act, which this Committee recently advanced. These reforms will help providers deploy fiber faster and farther so that rural and underserved communities are not left waiting for the infrastructure that will power America's digital future. Frontier AI and Critical Infrastructure. Frontier AI models hold vast potential to revolutionize networking. When deploying models with advanced cyber capabilities, the entire ecosystem needs to be prepared. Providing critical infrastructure like telecom with access to advanced models can help communications providers patch vulnerabilities on the network--and actively defend and respond to attacks. Question 1. Mr. Spalter, we've seen frontier labs like Anthropic and OpenAI release models with advanced cyber capabilities to selective groups of companies. Do you think companies that run communications networks should be part of those early cohorts? Answer. Yes. Telecom providers are actively strengthening their networks against AI-enabled threats by deploying advanced detection, vulnerability management, and automated response capabilities. We are also working closely with CISA, ONCD, NSA, Treasury, and NIST's Center for AI Standards and Innovation to share threat information, evaluate emerging capabilities, and translate frontier model insights into practical defenses for critical infrastructure. Collaboration is facilitated by safe and secure information sharing between government agencies and the private sector. Congress must reauthorize CISA 2015 in order to ensure secure information sharing continues between the government and industry. Question 2. What is the risk of letting the communications sector lag behind in access to advanced defensive technology? Answer. Threat actors are already using artificial intelligence to automate reconnaissance, accelerate vulnerability discovery, improve phishing campaigns, and increase the speed and sophistication of cyberattacks. If defenders do not have access to equally advanced defensive capabilities, the balance shifts in favor of attackers. Ensuring communications providers have responsible access to advanced defensive AI can help accelerate vulnerability identification, improve threat detection, automate incident response, and strengthen network resilience. As with information sharing under the Cybersecurity Information Sharing Act of 2015 and USTelecom's work with NIST on the AI Cybersecurity Framework Profile, the goal should be to ensure that America's critical infrastructure operators have the tools they need to defend the Nation's networks while maintaining appropriate safeguards and human oversight. ______ Response to Written Questions Submitted by Hon. John Hickenlooper to Jonathan Spalter Broadband Equity Access and Deployment Program (BEAD) Delays There is a widely recognized need for advanced communications infrastructure, especially for networks that are used every day by families, small businesses, schools, and healthcare facilities. Yet, in the past year the NTIA has tremendously slowed down the implementation of the BEAD program. They continue to change guidelines, creating a concerning level of uncertainty for service providers and their customers. In Colorado, the latest guidelines meant that fewer households were deemed as underserved, reducing how many Coloradans will be reached. BEAD funding was cut from $826 million to $420 million which will serve about 20,000 fewer locations. Question 5. Mr. Spalter, what impacts have these delays had on connecting rural and underserved communities? Answer. Delays in broadband deployment mean that rural and underserved communities must wait longer to access the high-capacity connectivity needed to participate fully in today's economy and to benefit from AI-enabled services. Laying fiber as fast and as far out into the network as possible is our best opportunity to deliver the next generation of American innovation to our customers and your constituents. AI will place unprecedented demands on communications networks, requiring substantial upload capacity in addition to download capacity. Fiber is the only broadband technology that provides the symmetrical capacity, scalability, low latency and reliability needed to meet the long-term demands of AI, quantum networking, and future innovations. The biggest impediment to achieving that goal is permitting. Delays and uncertainty in the deployment process slow construction, increase costs, and postpone the economic, educational, healthcare, and public safety benefits that broadband infrastructure makes possible. That is why USTelecom supports legislation such as the CLOSE THE GAP Act, and the RAIL Act, and the as well as the Accelerating Broadband Permits Act, which this Committee recently advanced. These reforms will help providers deploy fiber faster and farther so that rural and underserved communities are not left waiting for the infrastructure that will power America's digital future. Restructuring of BEAD Program In the past year, NTIA has completely restructured the BEAD program in which the Benefit of the Bargain rollout has mandated that states pick the lowest-cost bid that met performance standards. This prioritizes cheaper wireless and satellite alternatives that may sacrifice long-term network capacity and reliability, and at times do not adequately meet the needs of AI. In addition to a shift away from fiber, another component of this reconstruction is new rounds of rebidding to qualify for the program. Question 6. Mr. Spalter, what types of challenges arise for smaller Internet service providers when they are asked to rebid? Answer. For local providers in particular, repeated rebidding diverts limited resources away from construction and toward unplanned administrative requirements as well as delays. BEAD applications come with substantial costs as the process requires significant investments of time and capital, including network engineering, permitting analysis, and rescheduling construction crews. The cost of providing service in rural areas is already higher than urban areas so local providers often operate with streamlined budgets and limited personnel resources. When required to rebid there are already lost sunk costs so they are forced to decide whether they can afford to move forward on a rebid or withdraw from the process altogether--leaving a community without fast, reliable broadband internet. ______ Response to Written Question Submitted by Hon. Edward Markey to Bob Everson Question 1. Many emerging AI powered devices, from household robots to connected equipment used by businesses, will not have sim cards so cannot connect directly to a mobile network. Instead, they will likely connect over Wi-Fi, as most connected devices do today. What role will Wi-Fi play in supporting the next generation of AI powered devices, and what should policymakers do to ensure Wi-Fi has the capacity and spectrum needed to support continued AI growth? Answer. Wi-Fi plays a critical role in connecting devices used in home and business environments where significant traffic volume takes place, and where high throughput/low-latency connections matter most. As we move toward the ``Era of Physical AI,'' Wi-Fi will play a significant role in connecting these devices because it is cost- effective, ubiquitous, and perfectly suited for the local traffic that AI agents and AI enabled devices generate. As noted in my written testimony, we expect a 209 percent increase in campus and branch traffic over the next 3 years. Half the respondents to a recent survey of 3,400 IT professionals across 15 countries stated that their AI-driven demand was concentrated in their campus Wi-Fi networks. To ensure American consumers and businesses are fully able to enjoy the benefits AI will bring, it is critical that we preserve and expand use of the full 6 GHz band. Additionally, several Physical AI use-cases--like smart manufacturing facilities and ``self-aware'' buildings--would benefit from higher power operations within the 6 GHz band. These rule changes would allow Wi-Fi to reach further and handle the massive data payloads required by AI-driven video analytics and robotics. ______ Response to Written Questions Submitted by Hon. John Hickenlooper to Bob Everson Cybersecurity Risks of AI Agents AI agents introduce new cybersecurity risks that existing security practices do not fully address. 2026 reporting has revealed that only 59 percent of telecommunications companies say they have a robust methodology for identifying, assessing, and mitigating the risks associated with AI. Question 1. Mr. Everson, what are the most effective ways in which providers can prepare for increasingly autonomous AI systems managing portions of their networks? Question 2. Mr. Everson, what do you see as the most difficult barriers that providers face in responding to this ever-evolving cyberthreat environment? Answer. As noted in my written testimony, Cisco does not operate our customers' networks. However, there are a few ``non-negotiable'' cyber hygiene practices we encourage all network operators to take. These practices take on added urgency in the era of AI-fueled attacks operating at the speed of machines and the scale of networks, regardless of which critical infrastructure sector they are in: Patch and Upgrade. As I noted in my testimony, the volume of vulnerabilities that will be discovered by AI is increasing significantly--as are defenders' abilities to develop patches and upgrades and to help network operators test and deploy them using technology like AI-enabled digital twins. Remove/Replace ``End-of-Life'' (EOL) equipment that is no longer supported and cannot be patched or upgraded. Protecting and shielding. In some cases, it is impractical to immediately remove EOL equipment due to cost, complexity, or mission-readiness/up time. While it is ideal to patch/update devices capable of being supported and removing devices too old to secure, more focus and attention should be aimed at protecting or shielding technology that cannot be readily patched or pitched near term. AI-powered innovation can help bridge the gap by providing compensating controls while patches are being developed, tested, and deployed or until such time as unsupported technologies can be effectively swapped out for more modern, secure generations. Use phishing-resistant multi-factor authentication (MFA) for all accounts, not just administrative accounts. Implement Zero Trust Network architectures and least privileged access to prevent lateral movement within networks. Implement logging and monitoring for signs of malicious or unauthorized activities. Telecommunications networks are among the largest and most complex networks on Earth, and no two are alike. Ensuring network security in the AI era requires moving at machine speed. Leveraging AgenticOps to improve visibility into the network, manage complexity, and implement AI-enabled cybersecurity practices--in addition to the non-negotiable practices above--will help network operators respond to the fast- evolving threat landscape in the AI era. As with most technology transitions, the barriers I see among customers around the world are cost, complexity, and workforce. ______ Response to Written Questions Submitted by Hon. Maria Cantwell to Asad Ramzanali Dangers of AI. Just weeks ago, we saw another example showcasing just how rapidly AI models are advancing, when one of OpenAI's advanced models went rogue, breached its sandbox, and hacked into Hugging Face. And experts have warned for years about the ability of AI models to design bioweapons and even novel pathogens that we would not be able to detect. I am all for innovation and ensuring that America is the leader in artificial intelligence. But the fact remains: as AI models become more and more advanced, they can present complex, often unchecked, security risks. That is why we need our scientists and national security experts to test advanced models and mitigate their risks. Consumers, businesses, and communities must have assurance that AI is safe to use, reliable when it matters most, and won't endanger our national security. Question 1. How have recent security incidents like Hugging Face exposed vulnerabilities and risks to consumers? Answer. The recent OpenAI-Hugging Face incident and associated breaches, along with similar incidents later reported by Anthropic, show that AI models are becoming increasingly capable of strengthening cyberattacks and that much of our digital infrastructure contains vulnerabilities that, when exploited, can affect consumers, businesses, critical infrastructure, and government. If Hugging Face's systems can be breached by a model in testing, a motivated and well-financed adversary equipped with sophisticated AI-enabled cyberattack capability could reasonably be expected to cause real damage in the digital systems that undergird critical infrastructure and core parts of modern society. Banks, hospitals, water systems, airlines, electrical grids, and telecommunications networks are just a few of the types of systems that run on software of varying age and quality where advanced AI-based cyberattack scenarios could significantly degrade our day-to-day lives. Breaches to those systems can mean diverted funds, delayed surgeries, threats to water safety, canceled flights, energy outages, and surveilled calls. We should not be willing to tolerate living in a world where these systems fail consumers or are down for hours, days, or weeks because of a barrage of AI-enabled cyberattacks, though I fear that's our direction of travel. As I previously outlined, I believe the right policy response has three components. First, we must harden the cyber defenses of digital systems used to operate critical infrastructure, in government, and by other large organizations. This should include providing structured access to leading AI models, before they're deployed. Second, advanced AI models should undergo mandatory pre-deployment testing. This should apply to closed-weight and open-weight models alike. However, the techniques to mitigate vulnerabilities in testing should be different to account for the distinct technical affordances of closed-and open- weight model ecosystems. A testing requirement should not be a ban on open-weight models because they play a critical role for competition, research, and security. Third, a governmental body must thoroughly investigate incidents like the OpenAI-Hugging Face breach to establish facts, distill lessons, and provide companies and the public with a clear accounting of what happened and how it'll be avoided in the future. Question 2. What is the role of our national labs and Federal resources when it comes to testing AI systems? Answer. The Federal government should use all of the expertise and authorities at its disposal to test models, but also to inform pre- deployment requirements and investigate incidents. The Department of Energy's National Labs employ many world-leading scientists and technologists who should contribute to the technical aspects of developing and operating AI model testing. The Labs have decades of experience supporting national security missions, and AI model testing falls squarely within those missions. For example, the Labs' expertise and experience with chemical, biological, radiological, and nuclear (CBRN) weapons technologies, and how those technologies intersect with AI, should inform any testing regime. AI Preemption. My policy approach is clear--we need a strong Federal AI framework that protects consumers, workers, children, and creators while sustaining innovation and U.S. technological leadership. But until we set a clear Federal standard, we cannot and should not preempt state AI laws. Preemption leaves a vacuum. Preempting state AI laws before a Federal framework exists leaves consumers with no protection at all. In fact, states are currently the only actors policing AI harms. Question 3. How are states filling in the gaps to protect consumers in the absence of a Federal AI standard? Answer. State laws are among the only legal protections Americans have from AI's harms today, and Congress should not preempt state statutes without commensurate Federal protections. Illinois's legislature made the use of AI in employment contexts (e.g., hiring, disciplining) illegal if the system discriminates against protected classes. Connecticut curbed surveillance pricing, where automated systems (including AI-based ones) are used to develop personalized prices based on individual data. While these laws may not be perfect, they are examples of real state laws protecting real Americans from real harms of AI systems. Congress should not undo these efforts without commensurate, legally enforceable consumer protections. AI and Network Security. In 2024, we saw the worst telecom hack in our Nation's history when Chinese government-sponsored hackers, known as Salt Typhoon, deeply penetrated the networks of at least nine U.S. telecommunications companies. The result left millions of Americans' phone, text, and location information exposed. The sophistication of frontier AI models--including the reports of the rogue attack on Hugging Face--only further exposes the security risks of these unchecked capabilities. Without meaningful guardrails, AI will continue to exploit vulnerabilities that could lead to the next sophisticated cyberattack on our networks. Question 4. Given what we've seen recently with rogue AI models exploiting network vulnerabilities, could a more sophisticated hack like Salt Typhoon happen again? Answer. Salt Typhoon's hacks occurred before advanced AI models started bolstering cyberattacks. The group used stolen passwords and known vulnerabilities associated with telecommunications network equipment. The result was disastrous. Chinese state-backed hackers had the ability to intercept calls and texts from high-profile Americans, including presidential candidates. This was a major breach of national security and consumer privacy. AI models are now able to identify zero-day vulnerabilities, chain exploits, and act at machine speed against many targets simultaneously in ways that could enable far worse cyberattacks on our telecommunications networks than what Salt Typhoon was able to do just a couple of years ago. I worry deeply about the threats posed by AI- enabled systems (e.g., AI agents) that have the capability to constantly look for and act on network vulnerabilities. Question 5. How can consumers be confident that their providers will deliver secure, reliable communications when they need it most? Answer. In an ideal world, consumers should generally not have to think about cybersecurity. For decades, companies, cyber professionals, and even policymakers have described cyber incidents as the outcome of individuals using weak passwords or failing to patch software, rather than viewing the actual problem as digital systems that allow weak passwords or that are vulnerable to hacks of individual accounts turning into systemic failures. We have to view cybersecurity more as a systems issue and place duties and burdens on the entities most able to absorb those responsibilities, starting with digital infrastructure providers. Consumers should have confidence when companies that collect personal information or that control critical systems are required to protect that information and those systems, when independent authorities accountable to the public have approved underlying security measures, and when consumers have recourse in scenarios where they are harmed. Future-proof Networks. We need safeguards to protect consumers from AI's risks, but we also need to build the infrastructure that makes AI's benefits real for all Americans. Artificial intelligence depends on high-speed, low-latency connectivity, especially fiber to homes, businesses, and anchor institutions. Yet this Administration has spent eighteen months slowing broadband deployment to a crawl, steering funding away from fiber toward slower, less reliable connections, and sitting on guidance states need to prepare their networks and communities for AI. Earlier this month, NTIA Administrator Arielle Roth gave a speech titled ``There Is No Such Thing as Future-Proof Technology,'' arguing government shouldn't favor fiber because it could be leapfrogged. Question 6. Mr. Ramzanali, you are an expert on these issues--do you think this Administration's strategy on broadband deployment is helping or hurting to maintain U.S. leadership on AI? Answer. American leadership in AI depends on all Americans having access to AI. Fiber networks let households, businesses, and anchor institutions access AI tools, and they move data in and out of AI data centers. Despite Congress passing a law that makes clear that all Americans deserve access to the highest quality Internet connectivity and providing $65 billion to that end through BEAD, the Middle Mile Broadband Infrastructure Program, and other programs authorized under the Bipartisan Infrastructure Law, this administration has delayed deployment and introduced ambiguity for all involved. These actions hurt American leadership in AI. The implication of these actions, and of NTIA's recent public statements on fiber, is that rural and low-income Americans get left behind. To be clear, fiber demonstrably and consistently delivers higher speeds, greater bandwidth, lower latency, and higher reliability. That's what was needed before the AI boom, and it's what's needed now. However, more than NTIA statements, I worry about the administration's actions and inactions. Congress passed a law that prioritizes speed, scalability, reliability, and low latency, which fiber uniquely and consistently delivers. In that context, NTIA delaying the implementation of BEAD and deprioritizing fiber should be seen as sidelining rural and low-income Americans, postponing their access to jobs and education that require broadband, and failing to implement a law passed on their behalf. Frontier AI and Critical Infrastructure. Frontier AI models hold vast potential to revolutionize networking. When deploying models with advanced cyber capabilities, the entire ecosystem needs to be prepared. Providing critical infrastructure like telecom with access to advanced models can help communications providers patch vulnerabilities on the network--and actively defend and respond to attacks. Question 7. Mr. Ramzanali, do you agree that the frontier labs and the government need to work together to ensure that critical infrastructure operators, like the communications industry, have access to advanced defensive technologies? Answer. Yes, I agree that critical infrastructure providers should have access to advanced AI models before they are released to patch vulnerabilities exposed by the testing of those models and to bolster their cyber defenses more generally. Access should be managed by an explicitly authorized Federal program and not by ad hoc company decisions or White House processes. Question 8. What are the risks if we leave them behind? Answer. The ultimate risk of not enabling critical infrastructure providers to harden their cyber defenses is leaving consumers, businesses, government agencies, and others that depend on critical infrastructure vulnerable to cyber intrusions from bad actors and adversaries. More plainly, the foreseeable risk is that we are all more vulnerable to constant disruptions and degradation of critical infrastructure. The more worrying risk is that critical infrastructure systems fail in ways that cause damage to life, property, and national security. ______ Response to Written Questions Submitted by Hon. John Hickenlooper to Asad Ramzanali AI-Generated Fraud. Fraud has long been a challenge in the telecommunications industry, causing losses for consumers and companies. Mobile money and payment service banking introduce an even more complex environment, in which the risks significantly increase in frequency and impact. AI in the fraud landscape plays a dual role in that it can be used as both a threat and a shield. An example of this is the AI-generated voice cloning that has made it much easier to trick people with sophisticated robocall scams. Question 1. Mr. Ramzanali, beyond existing caller authentication tools like STIR/SHAKEN, what additional technical capabilities or authorities would most improve carriers' ability to detect AI-generated fraud? Answer. Despite several FCC proceedings and Congressional actions, robocalls remain a serious problem. Americans still receive billions of robocalls per month, and those cause billions of dollars of damage in the form of consumer scams. It is clear that industry actions are not working, and policy actions taken over the better part of a decade have not been commensurate with the scale of the problem. AI-generated voice cloning has made scams harder to detect for consumers and providers alike. However, instead of focusing on detecting or combating fraudulent audio, which will only become more challenging to detect, I recommend focusing on the parts of the network that have the capacity to mitigate robocalls. Before AI voice cloning was a major issue, public interest groups had long recommended systemic solutions to dealing with robocalls. For example, in June 2022, the National Consumer Law Center (NCLC) and the Electronic Privacy Information Center (EPIC) issued a report titled ``Scam Robocalls: Telecom Providers Profit,'' which recommended requiring all network providers to stop known illegal robocalls (where today they are permitted but not required to do so). Originating, gateway, intermediate, and terminating telecommunications network providers all have technical capacities to block illegal calls and should be required to use those capabilities. As the report further recommends, providers that knowingly transmit illegal robocalls should face financial penalties. STIR/SHAKEN is a good example of an important but insufficient tool. It is a protocol that provides information about whether a call is coming from a caller authorized to use the number originating the call. Requiring providers to adopt this protocol simply requires them to receive and transmit information. We need to go beyond information sharing. Restructuring of BEAD Program. In the past year, NTIA has completely restructured the BEAD program in which the Benefit of the Bargain rollout has mandated that states pick the lowest-cost bid that met performance standards. This prioritizes cheaper wireless and satellite alternatives that may sacrifice long-term network capacity and reliability, and at times do not adequately meet the needs of AI. In this shift, NTIA has also moved away from prioritizing fiber, against recommendations from industry experts. Question 2. Mr. Ramzanali can you speak to the need for fiber in managing AI's expected growth? How are these policies misaligned with the needs you are currently seeing? Answer. American leadership in AI depends on all Americans having access to AI. Fiber networks let households, businesses, and anchor institutions access AI tools, and they move data in and out of AI data centers. Despite Congress passing a law that makes clear that all Americans deserve access to the highest quality Internet connectivity and providing $65 billion to that end through BEAD, the Middle Mile Broadband Infrastructure Program, and other programs authorized under the Bipartisan Infrastructure Law, this administration has delayed deployment and introduced ambiguity for all involved. These actions hurt American leadership in AI. The implication of these actions, and of NTIA's recent public statements on fiber, is that rural and low-income Americans get left behind. To be clear, fiber demonstrably and consistently delivers higher speeds, greater bandwidth, lower latency, and higher reliability. That's what was needed before the AI boom, and it's what's needed now. However, more than NTIA statements, I worry about the administration's actions and inactions. Congress passed a law that prioritizes speed, scalability, reliability, and low latency, which fiber uniquely and consistently delivers. In that context, NTIA delaying the implementation of BEAD and deprioritizing fiber should be seen as sidelining rural and low-income Americans, postponing their access to jobs and education that require broadband, and failing to implement a law passed on their behalf. ______ Response to Written Questions Submitted by Hon. Brian Schatz to Asad Ramzanali Fiber. For Hawaii, interconnection of first, middle and last mile fiber networks is key; not only to ensure the delivery and arrival of data, but also for the resiliency of our multi-island network. Can you elaborate on the importance of these fiber networks in powering AI research and infrastructure? Can you elaborate on the role of historic broadband investments, such as BEAD, in achieving these goals? Answer. American leadership in AI depends on all Americans having access to AI. Fiber networks let households, businesses, and anchor institutions access AI tools, and they move data in and out of AI data centers. Despite Congress passing a law that makes clear that all Americans deserve access to the highest quality Internet connectivity and providing $65 billion to that end through BEAD, the Middle Mile Broadband Infrastructure Program, and other programs authorized under the Bipartisan Infrastructure Law, this administration has delayed deployment and introduced ambiguity for all involved. These actions hurt American leadership in AI. The implication of these actions, and of NTIA's recent public statements on fiber, is that rural and low-income Americans get left behind. To be clear, fiber demonstrably and consistently delivers higher speeds, greater bandwidth, lower latency, and higher reliability. That's what was needed before the AI boom, and it's what's needed now. However, more than NTIA statements, I worry about the administration's actions and inactions. Congress passed a law that prioritizes speed, scalability, reliability, and low latency, which fiber uniquely and consistently delivers. In that context, NTIA delaying the implementation of BEAD and deprioritizing fiber should be seen as sidelining rural and low-income Americans, postponing their access to jobs and education that require broadband, and failing to implement a law passed on their behalf. Cybersecurity. What specific hardening measures do you recommend for telecommunications networks to strengthen network cybersecurity posture? Are current Federal requirements for network security sufficient to mitigate AI agent-assisted cyberattacks? Does the emergence of AI agent-assisted cyberattacks warrant new considerations? Answer. AI models are now able to identify zero-day vulnerabilities, chain exploits, and act at machine speed against many targets simultaneously in ways that could enable far worse cyberattacks on our telecommunications networks than what Salt Typhoon was able to do just a couple of years ago. Current requirements for network security are not sufficient to mitigate AI agent-assisted cyberattacks. The FCC could mandate several measures to harden telecommunications networks. Example actions include: implementing clearly defined and narrowly prescribed role-based access controls; preventing unauthorized interception; eliminating default credentials; requiring multifactor authentication for most systems; mandating software patch timelines for internal and contractor-managed systems; replacing older network equipment where vendors no longer provide security updates; segmenting systems to avoid exploits in one system enabling bad actors to access other systems; applying the strongest measures to each system, not just core network systems; identifying systems used across the entire supply chain and requiring similar measures of them through contractual obligations; and regularly auditing and testing systems. None of these concepts are new to telecommunications providers. What's needed is for the FCC to mandate these and other measures and enforce against providers when they fail to take these actions. Financial Risks from AI buildout. What financial risks does the AI buildout pose to investments that impact everyday Americans, such as retirement? Are there any types of investments that are particularly exposed? How much of a systemic risk does the AI buildout pose to our financial system? Is the extent of this risk easily measurable? Answer. Today's AI buildout poses significant risks to Americans' retirement accounts, the broader financial sector, and the whole economy. Retirement accounts are often diversified across various asset classes. Almost all of those asset classes are dependent on AI today. The Magnificent Seven tech company stocks account for over a third of the S&P 500, and they all depend on the AI boom continuing. Record corporate bond issuances and private credit lending, along with investments from life insurance and pension funds, are also now highly dependent on AI. A large majority of GDP growth since at least early 2025 has depended on AI investments. However, the troubling magnitude of these investments is only one half of the problem. The other part is the interconnected, circular financial engineering involved. Chip companies, hyperscalers, and AI companies are each other's customers and vendors, competitors and partners, and investors and investments. They are increasingly depending on unregulated private credit deals, shifting risk off their books to special purpose vehicles (SPVs), offloading computing to highly levered neocloud vendors, and accessing capital from credit default swaps and asset-backed securities. This means a problem at one company could cascade to the whole sector and to the entire financial system. We have recent precedent for how things might unfold. When the dot- com bubble burst, thousands of companies went under, hundreds of thousands of people lost jobs, and millions of Americans collectively lost trillions of dollars. A global recession followed. However, as I've written in a recent report entitled ``After the AI Crash,'' I fear today's AI financial realities more closely resemble the conditions that led to the 2008 financial crash. We understand many aspects of the risks of AI financing--enough to start taking actions, including untangling the web of financial engineering. However, many questions remain, and the opacity of the financial engineering involved is itself a major risk. Private credit lacks transparency; SPVs hide debt off the books of public companies and evade disclosure requirements; derivatives entangle investors beyond those close to AI. Studying, quantifying, communicating, and mitigating the systemic risk should be a top priority of financial regulators, the Federal Reserve System, and the White House. [all]