Bipartisan ‘Great American AI Act’ Draft Would Freeze State AI Laws for Three Years

A pair of House lawmakers on June 4 unveiled a long-awaited federal framework for artificial intelligence, releasing a 269-page discussion draft of the Great American Artificial Intelligence Act of 2026. The bipartisan effort is led by Rep. Jay Obernolte (R-Calif.) and Rep. Lori Trahan (D-Mass.), with co-sponsor Rep. Erin Houchin (R-Ind.), and is being circulated to gather feedback before any formal introduction.

The draft is built around four pillars: establishing governance for frontier AI models, gathering data on how AI is reshaping the U.S. workforce, strengthening cybersecurity, and spurring new AI research and development. Its most contentious provision would preempt state laws that specifically regulate the development of frontier AI models, with a three-year sunset. Crucially, that preemption would not reach laws governing how AI is used or deployed, and states could still pass rules of "general applicability" — a carve-out the sponsors stress, but one that critics say leaves a major gap.

In exchange for that federal floor, the bill imposes real obligations on the biggest labs. Large frontier developers — those above $500 million in annual revenue — would have to publish safety frameworks, report critical safety incidents to the government, and submit to semi-annual third-party audits. The draft also formally codifies the Commerce Department's Center for AI Standards and Innovation (CAISI), authorizing $100 million a year for fiscal 2027 through 2029, as reported by Roll Call and Nextgov/FCW.

Beyond model governance, the text is sweeping: it directs the Government Accountability Office to evaluate federal AI adoption, creates penalties for AI-based impersonation of government officials, revises Census Bureau and Bureau of Labor Statistics surveys to track AI's labor-market effects, sets up a public-private testbed for evaluating AI systems, pushes international standards leadership through NIST and the Energy Department, funds CISA grants for open-source software security, and codifies the National AI Research Resource. Rep. Houchin framed the stakes bluntly: "America should lead the world in artificial intelligence, not regulate ourselves into falling behind China."

The three-year preemption immediately drew sharp criticism from AI-safety groups, who argue it would handcuff states such as Colorado and California just as their own AI laws take effect. For BitsMinds readers, the draft is a striking counterpoint to the same week's news abroad: where Canada's new "AI for All" strategy leans on public investment and adoption and the EU's AI Act emphasizes binding rules, Washington's emerging approach trades a temporary federal ceiling on state regulation for transparency obligations on the largest developers. With the bill still a discussion draft, the fight over that trade-off is only beginning.