House Bipartisan Bill Restricts State AI Regulation

Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) on Thursday released a discussion draft of the Great American Artificial Intelligence Act, a bipartisan attempt to set a national framework for AI, according to Axios and The Hill. Axios describes it as a 269-page framework and lists co-sponsors including Reps. Scott Franklin (R-Fla.), Suhas Subramanyam (D-Va.), Erin Houchin (R-Ind.), and Scott Peters (D-Calif.). The Hill, which obtained the draft, reports the lawmakers describe it not as a "final product" but as a way to gather feedback before any formal introduction.

The Hill reports the draft would override state regulations targeting AI model development for three years, while a summary of the bill states it "expressly does not preempt laws of general applicability, common law remedies, or laws regulating AI use or deployment." Both Axios and The Hill report the bill would codify the Center for AI Standards and Innovation (CAISI), which The Hill notes was launched in June 2025 under the Commerce Department. Axios reports CAISI would set voluntary standards and guidelines and receive $100 million per year from 2027 to 2029; The Hill adds it would license independent organizations to audit lab compliance. The draft also covers AI whistleblower protections, higher penalties for AI-enabled fraud, AI literacy and research funding, and federal data collection on AI's labor-market impact, including an AI Workforce Research Hub overseen by the Secretary of Labor, per Axios and The Hill.

Under the draft, The Hill reports, "large frontier developers" would have to write and implement plans to address risks before releasing new models and publish a "frontier AI framework" describing the technical and organizational protocols they use to evaluate and manage a model's "catastrophic risks." Developers would also be required to file a report with CAISI shortly after certain safety incidents, including those posing "imminent risk of death or serious injury," according to The Hill.

The Hill reports the draft is Congress's first substantive response to a White House AI policy "wish list" and arrives days after President Trump signed an executive order creating a voluntary government testing process of up to 30 days for unreleased AI models, down from an earlier 90-day proposal. In a Bloomberg Law op-ed cited by The Hill and Axios, Obernolte and Trahan wrote that the framework "must be durable enough to survive changes in Congress, administrations, and political priorities" and argued that "rather than allow protections to exist only in a handful of states or force innovators to navigate dozens of different legal regimes, our framework would establish one national standard." The preemption provision is contested: consumer-protection group Public Citizen argues the bill would strip states of authority "to protect consumers, workers, and children."

Federal-first approaches that pair a single national standard with mandatory disclosure, risk planning, and independent audits resemble supervisory models used in other regulated technology domains, and codifying an existing standards body rather than creating a new regulator is a common way to centralize technical review. Whether a federal floor that preempts state rules simplifies compliance or weakens accountability is a long-running tension in technology policy, and the outcome here will depend on which provisions survive public comment.

Track whether the three-year preemption window is narrowed or removed amid pushback, how CAISI's authority and the "frontier AI framework" disclosure rules are defined, and whether the audit and incident-reporting requirements align with existing industry practice for model documentation and red-teaming. The lawmakers have framed the draft as "the start of a serious national conversation," so substantial changes are likely before any formal bill.