The 2026 midterm elections delivered what many have called a “blue tidal wave” at both the federal and state levels. As of November 1, Democrats had won 256 House seats to Republicans’ 169 and 55 Senate seats to Republicans’ 44. Democrats also won a majority of gubernatorial races. Many factors appear to have driven the results, but artificial intelligence (AI) stood out as a newly prominent issue with candidates from both parties campaigning on concerns about AI-related job losses, rising utility bills viewed as linked to data centers, and children’s safety as America’s youth spend increasing amounts of time engaged with AI chatbots. AI is poised to top the new Congress’s agenda in January. This client alert predicts five key trends that will affect companies developing or deploying AI and provides steps they can take now to prepare.
#1 Congressional oversight of AI policy, and policymakers, will increase – alongside a parallel focus on companies’ safety practices.
New Democratic majorities will use hearings, document requests, and subpoenas to examine and expose how executive branch agencies are addressing AI policy concerns, including export controls, war planning and execution, and cybersecurity.
- Expect more Commerce Department scrutiny on export decisions for advanced AI chips, particularly the risk of diffusion and Chinese espionage. Democrats have already opened an investigation into Commerce’s approval of chip sales to China, amid reports that the department paused a probe into suspected chip smuggling. We anticipate that lawmakers will press officials on how those licensing decisions were made, whether political or industry pressure played a role, and how Commerce enforces requirements that certain exported chips not be resold or further transferred to restricted users. Lawmakers may also look closely at the Center for Advancing Innovation and Standards for Super Intelligence (CAISSI), formerly the U.S. AI Safety Institute. They are likely to ask how its mission has changed since it was (twice) renamed, what model evaluations and standards work it is doing now, and how recent leadership and staffing changes have affected its capacity and independence.
- At the Department of War, lawmakers are likely to examine how AI is being used in military planning, what limits govern the use of AI, and how the department’s relationships with frontier developers has influenced procurement and other decisions.
- At the Cybersecurity and Infrastructure Security Agency (CISA), we expect questions about how the agency is countering the surge of AI-enabled cyberattacks, including the status of its efforts to stand up an AI-specific information-sharing center with the private sector to address AI-related security threats across critical infrastructure sectors. Lawmakers will also probe how recent staffing and budget cuts have weakened the agency’s ability to detect and respond to these threats.
We also expect closer scrutiny of frontier AI developers and deployers across industries. Lawmakers will likely focus, for example, on how companies test their models before release, decide how those models are used, and monitor them after deployment. We expect enhanced scrutiny regarding the hosting of Chinese open weight models. In addition, lawmakers may ask questions about bias testing, especially when that bias aligns with foreign government-sponsored narratives, potentially putting companies in a difficult position and intensifying a growing congressional focus on the outputs of leading AI models. Scrutiny may also extend beyond safety to companies’ financial ties to the Administration, including donations to the White House ballroom project.
As we explained here, companies should use the lame-duck period to identify potential areas of scrutiny, review previous government interactions, assess document-retention and response protocols, and prepare for inquiries beginning early in the new Congress.
#2 State and federal legislative activity will experience a dramatic uptick.
Despite an active debate over the last year, Congress is unlikely to pass a federal law broadly preempting state-level AI regulation. Democratic proposals have generally treated federal rules as a floor, not a ceiling, and have not entertained broad federal preemption of what now represents a growing body of state-level laws. Sen. Maria Cantwell (D-WA), who is positioned to chair the Senate Committee on Commerce, Science, and Transportation which has been leading AI-related initiatives, has opposed efforts to override state laws through a comprehensive federal AI law. States have also pushed back hard. In September, re-elected Minnesota Attorney General (AG) Keith Ellison joined a bipartisan letter from 26 AGs asking Congress to “prohibit[] the preemption of state laws and grant[] state officials full authority to enforce these federal protections [regulating AI].” Accordingly, states will likely continue to be the primary vector for laws regarding chatbots, deepfakes, algorithmic pricing, and data centers.
Nonetheless, we expect substantial federal legislative activity as lawmakers respond to constituents’ concerns about AI’s growing power and seek to show that they are taking AI-related risks seriously. We expect a range of federal AI bills on issues including frontier model safety, incident reporting, “kill switch” requirements, children’s safety on chatbots, algorithmic and surveillance pricing, AI-driven job displacement, the energy and utility costs of data centers, and tighter export controls on advanced AI chips. Several newly elected members built their reputations on state AI legislation and may try to repurpose their state-level solutions in federal law. For example, Representative-elect Scott Wiener (D-Calif.) authored SB 53, California’s frontier AI law (since replicated in New York and Illinois), which requires frontier developers to publish safety frameworks and report critical safety incidents. And Representative-elect Manny Rutinel (D-Colo.) sponsored SB 24-205, the first comprehensive state AI law (later repealed and replaced), which would have required risk assessments to guard against algorithmic discrimination in consequential decisions such as employment, housing, and lending. Moreover, it is possible that real-world developments—such as a high-profile, AI-enabled cyberattack—could accelerate the pace of legislative activity at both the federal and the state level and change the latitude lawmakers might have for more dramatic or sweeping types of legislation in this area,
Companies should closely track federal and state AI legislation and assess compliance costs in real time as bills advance, rather than waiting for enactment. Because federal rules will likely layer on top of state laws rather than preempt them, companies should map overlapping obligations and test their systems against leading proposals. Building a flexible compliance program around common elements (risk assessments, incident reporting, transparency, and human oversight) can reduce adaptation costs. Companies should also consider engaging early to help shape workable requirements.
#3 Expect scaled, fast enforcement by State AGs
As we explained here, the Democratic wave carried down the ballot to state AG races. Democrats flipped five Republican-held AG offices (Georgia, Iowa, Kansas, Ohio, and Texas) and are set to hold 29 AG offices to Republicans’ 22, with 11 new AGs taking office in January. State AGs were already among the most aggressive AI enforcers before the election, and we expect these investigations to grow in number and scope under traditional consumer protection and privacy laws.
Several incoming AGs made AI enforcement a campaign pledge:
- Texas AG-elect Nathan Johnson promised a 30-day audit of Texas consumer protection, competition, and employment laws for AI gaps, saying that “predatory practices are predatory, no matter what’s driving them.”
- Michigan AG-elect Eli Savit pledged: “During my first 100 days in office, I pledge to issue a suite of rules addressing the misuse of AI and then enforcing those rules.” His targets include algorithmic price-gouging, surveillance pricing, and deceptive marketing.
- Georgia AG-elect Tanya Miller and Nevada AG-elect Nicole Cannizzaro focused on pricing: Miller pledged to use state antitrust authority against companies that use AI and pricing algorithms to raise consumer costs, and Cannizzaro named data-driven personalized pricing and rent increases as priorities.
New state AGs often move quickly to deliver on campaign promises. Companies should use the transition period to proactively assess their exposure in states with new AGs, including near-term commitments. Companies should also revisit pending investigations and settlement talks, as new leadership in the AG’s office may take a different approach. Finally, companies should set up a process for AG inquiries, including document preservation and early involvement of counsel.
#4 Expect more bespoke AI standards, including NIST guidance
We expect the National Institute of Standards and Technology (NIST) to keep translating high-level AI risk-management principles into more specific technical guidance for emerging AI systems, including agentic systems, as it likely would have done regardless of the election outcome. We are watching two NIST workstreams in particular:
- Updates to the AI Risk Management Framework (we summarized the existing framework here), as directed by the 2025 AI Action Plan; and
- Forthcoming guidance on AI agents following the 2026 NIST AI Agent Standards Initiative to develop standards and guidance for the secure and interoperable use of autonomous AI agents.
NIST guidance is generally voluntary, but several state AI laws (for example, the Texas Responsible Artificial Intelligence Governance Act, or TRAIGA) are now tied to these standards. NIST alignment could therefore become an important practical compliance benchmark for companies developing or deploying AI. Further, there is a risk of tort liability for failing to comply with a reasonable framework, and NIST might provide a ready definition of reasonableness to courts looking for a standard. Two reelected Democrats also have NIST-focused bills to scale reliance on NIST framework. For example, Rep. Ted Lieu (D-Calif.) introduced the bipartisan Federal Artificial Intelligence Risk Management Act of 2026, which would require federal agencies to use the NIST AI Risk Management Framework and direct NIST to develop additional standards and resources. And Rep. Sarah McBride (D-Del.) sponsored the READ AI Models Act, which would direct NIST to develop standardized templates and technical guidance for documenting AI models and associated data.
Relatedly, we expect Congress to use its oversight authority to push other regulators to issue more tailored AI guidance. For example, Sen. Elizabeth Warren (D-MA) warned that recent federal banking policies could leave the financial system more vulnerable to AI-enabled cyberattacks by weakening examiners’ ability to enforce cybersecurity standards. Similarly, House Financial Services Democrats have pressed the Securities and Exchange Commission to explain how existing securities rules apply to AI agents that trade on behalf of retail investors, including what obligations brokers and AI developers bear and whether new legislation is needed. Further, Sen. Bill Cassidy (R-LA) released a discussion draft of the PATIENT AI Act, which directs the Department of Health and Human Services to issue guidance explaining how HIPAA-regulated entities should protect privacy and security when using AI, and clarify when AI-based technologies are subject to oversight by the Food and Drug Administration or the Office of the National Coordinator for Health Information Technology.
Companies should assess how their AI governance programs align with the NIST AI Risk Management Framework (both the current version and any future version), which may increasingly inform state-law obligations and what counts as “reasonable” care in tort claims. They should also be prepared for new AI guidance from sector regulators that have so far said little about AI, and consider engaging early to help shape that guidance.
#5 Industry partnerships will remain voluntary, though they may invite more scrutiny
We expect the Trump Administration to keep pushing AI companies to work together to identify risks and scale defensive measures as AI risks grow. This could include sharing threat information, coordinating safety and security testing, and developing tools to detect and respond to misuse. The emphasis will likely remain on voluntary cooperation rather than compulsory measures, consistent with the administration’s general aversion to heavy-handed AI regulation and to “safety” writ large.
As we described here, the administration’s current approach centers on Executive Order 14409, Promoting Advanced Artificial Intelligence Innovation and Security (June 2, 2026), which sets up voluntary government-industry coordination on frontier AI in place of mandatory licensing or preapproval. Under the order, the government designates “covered frontier models” through a classified evaluation process. Developers of these models may then engage with the government before release, provide limited advance access, and help identify “trusted partners” that receive early access to strengthen cybersecurity and critical infrastructure. The order also directs the establishment of an AI cybersecurity clearinghouse, which Treasury has implemented through its GOLD EAGLE initiative to coordinate AI-assisted vulnerability discovery and remediation with developers, CISA, and critical infrastructure operators.
Going forward, the key open question will be who gets access to frontier AI (and when) and how those decisions will be made. The order’s voluntary framework and trusted-partner list, which were due August 1, have not been released, and the threshold for covered models remains classified. Companies have reportedly been unsure whether access decisions rest with developers or the government. In August, Senators Kirsten Gillibrand (D-NY), Mark Warner (D-VA), Mark Kelly (D-AZ), Adam Schiff (D-CA), and Chris Coons (D-DE) pressed the Administration on the standards, legal authorities, decisionmakers, and appeal rights behind these decisions, but Treasury’s response has offered little detail to date. With Democrats now controlling both chambers, we expect oversight to focus on frontier AI model access, and the fairness and transparency around making these decisions.
Companies should think carefully before making voluntary commitments and should also confirm that those commitments are consistent with what they are telling Congress, regulators, and state AGs, as well as their obligations under state and local law. With Democrats now leading oversight and state AGs ramping up enforcement, any gap between a company’s voluntary pledges, public statements, and actual practices could become a basis for inquiry or liability. Companies should keep a clear record of what they commit to and how they meet those commitments.