Skip to Main Content
11/16/2026|10 minute read

Key Takeaways

  • Federal artificial intelligence (AI) policy is likely to divide along institutional lines in 2027. The Trump administration will continue to emphasize American technological leadership, infrastructure development and regulatory restraint, while the Democratic-controlled Congress is expected to focus more heavily on safety, accountability, consumer protection and the costs associated with AI. When the 120th Congress convenes in January 2027, AI policy is likely to look markedly different from the federal approach of the past two years.
  • Congressional oversight may move faster than legislation. New congressional leadership will control committee agendas, hearing calendars and subpoena authority, giving Congress the ability to compel testimony and documents from AI developers and other companies concerning model safety, cybersecurity incidents, testing, data practices and other AI risks.
  • Pending legislation identifies the controls Congress may expect, even if the bills do not become law. Legislative proposals from the current Congress provide a preview of what may come next: independent testing of frontier models, incident reporting, government oversight and human-controlled shutdown capabilities. These measures provide a useful indication of the governance practices likely to receive congressional attention in the next session. The bipartisan FRONTIER Act,[1] for example, would create a federal oversight regime for frontier AI and independent assessment of certain developers’ safety practices.
  • Regulatory fragmentation may persist. Presidential veto authority, Senate procedural requirements and continuing disagreements over the appropriate scope of AI regulation make enactment of a comprehensive federal statute uncertain. Companies should therefore prepare for overlapping federal oversight, executive action, state regulation, agency enforcement and private litigation rather than presume that a single national framework will emerge.

A. Two Federal AI Agendas

For much of the past two years, federal AI policy has largely been developed through executive action rather than comprehensive legislation. The Trump administration has emphasized innovation, American technological competitiveness, infrastructure development and reducing regulatory barriers to AI deployment. As we previously discussed, that approach culminated in a federal effort to challenge state AI laws viewed as inconsistent with a uniform, minimally burdensome national framework.

The 120th Congress may alter, but not displace, that approach. Although the executive branch will continue to direct federal agencies and enforce existing law, Democratic control of the House and Senate will give the party significant authority over the congressional agenda beginning Jan. 3, 2027. Perhaps most important, it will put congressional committees, investigative agendas, hearing calendars and subpoena authority under new leadership. Current reporting indicates that Democratic lawmakers are preparing proposals addressing advanced-model testing, AI-related harms and the costs associated with AI deployment and development.

Companies should therefore prepare for two federal AI agendas operating simultaneously: an executive branch continuing to emphasize innovation, technological leadership and regulatory restraint and a Congress placing focus on the risks associated with AI development and deployment. The resulting tension may shape federal AI policy even if Congress does not enact comprehensive legislation.

The distinction should not, however, be understood as a simple choice between regulation and deregulation. The executive branch itself continues to develop new mechanisms for addressing risks associated with increasingly capable AI systems, principally through enforcement against harms or unlawful conduct. In November, the Department of Justice (DOJ) announced a Super Intelligence Task Force focused on misuse of advanced technology by foreign and state-sponsored actors, transnational criminal organizations, and individuals or companies engaged in fraud or other unlawful conduct. The DOJ specifically identified regular reporting of risks and incidents and “vigorous internal controls” over advanced AI systems as areas of focus.

The emerging divide therefore may be less about whether AI should be governed than about how. Congress is more likely to explore prospective obligations concerning safety, testing, transparency and accountability, while the administration will continue to pair its support for innovation with enforcement against harmful or unlawful uses of advanced AI.

B. Congressional Scrutiny Will Carry Different Consequences

The change in congressional control will not necessarily produce entirely new areas of scrutiny. Democratic lawmakers have already used letters, information requests and other minority-party tools to examine AI companies and their practices. What will change in 2027 is the authority behind those inquiries. Committee chairs will have greater control over investigative agendas, hearings, witness testimony and compulsory process. As we previously discussed, this investigative authority is very broad and presents very different challenges than executive branch investigations and traditional litigation. A congressional investigation is not traditional litigation. Congress does not file a complaint. There are no Rules of Civil or Criminal Procedure or Rules of Evidence, and congressional investigative requests are not subject to the familiar discovery rules and proportionality requirements that govern litigation. Nor must a committee satisfy the same thresholds that constrain other forms of government investigations. For example, a congressional committee need not establish probable cause before issuing a subpoena, unlike federal law enforcement seeking a search warrant.

That distinction matters for companies that develop or deploy AI. An inquiry that begins with a voluntary request for information can develop into a broader investigation involving document production, executive testimony or subpoenas. And congressional investigations differ from traditional executive branch enforcement proceedings: They can develop rapidly, encompass conduct that is not itself unlawful and unfold publicly alongside legislative and political debate.

AI also presents an unusually broad range of potential subjects for congressional inquiry. Depending on the company and technology involved, committees could examine model testing and safety practices, reported incidents, cybersecurity controls, training and data practices, consumer harms, representations concerning AI capabilities or the adequacy of governance and escalation procedures. Congressional interest may also extend beyond model developers to companies that deploy AI in products, services, employment, financial, healthcare or other consequential settings.

Companies should therefore not wait for a subpoena to assess their readiness. Organizations facing potential congressional scrutiny should identify the internal personnel responsible for responding to congressional inquiries, understand where relevant records reside, review escalation and document preservation procedures, and consider whether public statements concerning AI safety and governance are consistent with internal policies and practices.

C. Frontier AI Proposals Are Beginning To Define Expected Controls

Recent bipartisan proposals concerning advanced, or frontier, AI systems provide an early indication of the safeguards Congress may expect from developers of the most capable models. Although the proposals differ in scope and approach, they increasingly focus on concrete operational controls, including independent testing, incident reporting, cybersecurity protections, risk monitoring and mechanisms for human intervention.

H.R. 9925, the “Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act,” or FRONTIER Act, introduced in July by Reps. Jay Obernolte, R-Calif., and Lori Trahan, D-Mass., with bipartisan cosponsors, would establish federal oversight of the development and deployment of frontier AI. Among other provisions, the legislation contemplates independent evaluation organizations assessing certain large developers’ governance, risk monitoring and mitigation practices.

Other proposals go further. For instance, H.R. 9917, the “AI Kill Switch Act,” introduced by Reps. Ted Lieu, D-Calif., and Nathaniel Moran, R-Texas,[2] would require covered entities to maintain technical capabilities to shut down certain covered technologies. Sen. John N. Kennedy, R-La., subsequently introduced S. 5417, the “AI Emergency Button Act,”[3] which likewise would require developers or operators of advanced AI systems to incorporate a human-controlled shutdown capability.

The incoming Congress may also broaden the range of frontier AI proposals under consideration. The incoming Senate will include members who campaigned on substantially more aggressive restrictions in the development of advanced AI. Democratic Sen.-elect Abdul El-Sayed of Michigan, for example, proposed pausing the training and release of the most powerful AI models until federal safeguards are established, prohibiting the development of superintelligent systems and recursive self-improvement, and creating a Cabinet-level agency empowered to inspect AI laboratories, block dangerous model releases and order shutdowns.

These proposals vary in scope and political viability but, when viewed together, reflect growing congressional attention to the independent evaluation of advanced models, the reporting of significant risks and incidents, safeguards against misuse or loss of control, and the circumstances in which the government might regulate the development and deployment of advanced AI systems. Even where Congress does not reach consensus on specific requirements, these issues are likely to shape the legislative debate in 2027.

D. Existing Laws May Matter as Much as New Legislation

Another significant question will be whether Congress seeks to create new liability rules for injuries attributed to AI. Current reporting indicates that Democratic lawmakers are considering proposals addressing when AI developers may be held liable for harms caused by their products. But companies should not assume that liability depends on enactment of an AI-specific federal statute. Existing consumer protection, antidiscrimination, privacy, intellectual property, product liability and other laws may already apply to conduct involving AI, depending on the technology, use and resulting harm.

Of course, that does not mean Congress will leave the liability framework untouched. Proposals such as the “Clear Liability for Artificial Intelligence Misconduct (CLAIM) Act”[4] from Rep. Trahan would create new federal remedies specifically addressing harms caused by AI systems. But the legislative debate over whether to create AI-specific causes of action will occur against an existing legal framework that regulators and private litigants are already using to address AI-related conduct.

The practical consequence is that companies may face liability risk even if broader federal AI legislation stalls. Organizations should therefore remain responsible for the accuracy, reliability, confidentiality and integrity of work produced with AI assistance. That principle is unlikely to change regardless of which AI bills ultimately become law.

E. The Future of a Uniform Federal AI Framework

Perhaps the most immediate consequence of the election for businesses operating nationally is what Congress may not do.

The Trump administration has pressed for a uniform federal approach that would limit or displace state AI requirements. The administration’s December 2025 Executive Order (EO) 14365, “Ensuring a National Policy Framework for Artificial Intelligence,”[5] expressly advanced a national AI framework, and the DOJ subsequently established an AI Litigation Task Force directed at challenging state laws considered inconsistent with federal policy.

On Sept. 29, the president issued EO 14434, “Inaugurating the Era of Super Intelligence.”[6] This EO directs executive branch departments and agencies to use “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI” in all official correspondence, public communications, websites, reports, policy documents and other non-statutory documents. Notably, the EO started a 60-day review period by the assistant to the president for science and technology to draft proposed legislation to establish a federal definition of SI.

With control of Congress changing hands in January 2027, broad federal preemption of state AI laws faces a substantially more difficult legislative path. That does not necessarily mean the federal-state conflict will subside. Instead, the forum in which that conflict plays out may shift. The executive branch can continue to challenge selected state laws while Congress considers federal standards that may not displace state requirements.

For businesses, this may be one of the most consequential developments of all. State AI requirements may continue to develop while Congress debates federal standards and the executive branch challenges selected state restrictions. Instead of quickly resolving the existing regulatory patchwork, the next two years could make the relationship between federal and state AI regulation even more complex. Companies operating across jurisdictions should therefore be cautious about dismantling state-specific compliance programs in anticipation of federal preemption. Until Congress acts or courts invalidate particular state provisions, existing state obligations remain relevant.

F. AI Infrastructure and Data Centers Enter the Regulatory Conversation

The congressional AI debate also is broadening beyond models, algorithms and data. Attempts were made to pass legislation before the Nov. 3 midterm election but were ultimately unsuccessful. On Sept. 16, the House of Representatives passed H.R. 9340, the “Ratepayer Protection Act,” by a vote of 417-3. On Sept. 30, a Senate effort to advance the bill fell short of the 60 votes required to invoke cloture on the motion to proceed by a vote of 57-43. The enormous computing demands associated with AI have placed data centers, electricity costs, grid capacity, permitting and infrastructure squarely within the policy debate. Democratic lawmakers are reportedly developing proposals intended to prevent residential customers from bearing increased utility costs attributable to data center construction and operation. And data center regulation emerged as a recurring campaign issue among members of the incoming Senate, frequently framed around affordability rather than technology policy alone.

Several senators-elect called for data center operators to bear their own electricity and grid upgrade costs, while others proposed restrictions on tax incentives, water use or new construction. Democratic Sen.-elect James Talarico of Texas, for example, proposed federal minimum requirements governing grid protection, environmental effects, jobs and transparency, together with requirements that large energy users pay their own costs. El-Sayed went further, calling for a pause on new data center development as part of a broader plan to restrict advanced AI development.

These campaign positions suggest that congressional scrutiny of AI infrastructure will focus increasingly on who bears the costs of deployment. Indeed, Congress is already considering broader permitting legislation with significant implications for data center construction and the infrastructure needed to support it. Accordingly, organizations making substantial investments in AI infrastructure should monitor congressional activity not simply for legislation expressly labeled as an “AI” bill. Some of the most consequential federal developments for AI deployment may emerge from energy, tax, cybersecurity, national security or infrastructure legislation.

G. What Organizations Should Do Now

The regulatory direction remains fluid, but companies need not wait for Congress to act before preparing.

Organizations developing or deploying AI should consider:

  • Reviewing AI governance and escalation procedures. Companies should be able to identify who is responsible for approving AI systems, monitoring material risks, escalating incidents and documenting significant decisions. Those procedures may matter not only for compliance with emerging legislation but also for enforcement risk. The DOJ’s Super Intelligence Task Force specifically identified risk reporting and “vigorous internal controls” as areas of focus.
  • Preparing for congressional scrutiny. Companies with significant AI operations should assess whether their internal documentation, safety testing, incident-reporting procedures and public representations are consistent and capable of withstanding scrutiny in a congressional investigation or hearing.
  • Mapping federal and state obligations separately. Organizations should resist assuming that a future federal framework will eliminate state requirements. Compliance programs should identify which obligations originate at the federal, state, contractual and sector-specific levels.
  • Reviewing AI vendor agreements. Contracts should address responsibility for testing, security incidents, data use, legal compliance, audit rights, indemnification and allocation of liability as regulatory requirements evolve.
  • Monitoring legislation beyond bills expressly regulating AI. Cybersecurity, energy, infrastructure, tax, procurement, national security, investor protection and consumer protection legislation all may materially affect companies developing or deploying AI.
  • Preserving flexibility. Perhaps most importantly, companies should avoid designing compliance programs around the assumption that today’s regulatory approach will remain static. The principal challenge over the next two years may not be any single AI rule but the interaction among congressional oversight, executive action, agency enforcement, state regulation and private litigation.

Conclusion

The 120th Congress is unlikely to settle the debate over how the United States should regulate AI. It may intensify it.

The Trump administration remains committed to a federal policy emphasizing AI innovation and American competitiveness, while incoming congressional leadership appears poised to place substantially greater emphasis on safety, accountability, liability and consumer protection. At the same time, state governments will continue developing their own approaches, and existing laws will continue to apply to AI even where Congress has not enacted technology-specific rules.

For companies, the practical lesson is straightforward: Do not wait for a comprehensive federal AI statute. The more immediate risks may come from congressional investigations, executive branch enforcement, evolving state requirements, existing regulatory authority, litigation and the rapidly changing expectations surrounding responsible AI governance. Organizations that can explain what AI they use, why they use it, how they assess its risks, who supervises it and what happens when something goes wrong will be better positioned for whatever framework ultimately emerges.

BakerHostetler’s White Collar, Investigations and Securities Enforcement and Litigation team, Artificial Intelligence group, Federal Policy team and Congressional Investigations team draw on experience across Congress, federal agencies, emerging technology and complex investigations to advise clients on government investigations, regulatory compliance, AI governance and emerging technology risks. Our teams include former members of Congress from both sides of the aisle, former congressional and senior committee staff, and experienced white collar defense lawyers who counsel clients responding to congressional inquiries, preparing for interviews and testimony, and assessing investigation and litigation risks. Please feel free to contact any of our experienced professionals with questions regarding these developments or their implications for your organization.


[1] https://www.congress.gov/bill/119th-congress/house-bill/9925/text

[2] https://www.congress.gov/bill/119th-congress/house-bill/9917?s=2&r=3

[3] https://www.congress.gov/bill/119th-congress/senate-bill/5417?s=3&r=5

[4] https://trahan.house.gov/news/documentsingle.aspx?DocumentID=3861

[5] https://www.whitehouse.gov/presidential-actions/2025/12/eliminating-state-law-obstruction-of-national-artificial-intelligence-policy/

[6] https://www.whitehouse.gov/presidential-actions/2026/09/inaugurating-the-era-of-super-intelligence/


Featured Insights