Congress Unveils Bipartisan Bill to Set Federal AI Rules — and Freeze State Laws for Three Years

A bipartisan pair of House lawmakers released a major discussion draft today that would, if passed, reshape how artificial intelligence is regulated across the United States — and temporarily strip states of their authority to write their own AI laws.

Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA), working from the House AI Task Force's findings, published the Great American Artificial Intelligence Act of 2026, a sweeping bill that takes direct aim at the fragmented patchwork of state-level AI rules that has emerged over the past two years.

What the Bill Does

The centerpiece is a three-year federal preemption of state laws specifically targeting the development of frontier AI models. States could still regulate how AI systems are used within their borders and pass laws of "general applicability," but they would be barred from imposing their own development standards on the most capable AI systems during that window.

The draft also formally codifies the Center for AI Standards and Innovation (CAISI) within the Commerce Department, authorizing $100 million in annual federal funding from fiscal years 2027 through 2029. CAISI would be tasked with developing voluntary guidelines, evaluating AI systems, and monitoring progress across the industry.

Frontier AI developers — broadly defined as companies training the most capable models — would face a set of new federal obligations under the bill:

• Report critical safety incidents to the federal government
• Publish AI governance frameworks and demonstrate compliance with CAISI standards
• Identify risk thresholds and assess whether their models pose "catastrophic risks"
• Maintain cybersecurity defenses for private model weights
• Submit to independent third-party audits every six months
• Disclose model release dates and modifications publicly

The bill also includes whistleblower protections for AI researchers and employees, fraud provisions targeting AI-generated impersonation of government officials, and a directive for federal statistical agencies to update their surveys to track AI adoption across the US economy.

The Preemption Fight

The three-year state preemption clause is already drawing sharp criticism. The ACLU, Public Knowledge, and Public Citizen have each pushed back, arguing the provision would block states from enacting new consumer protections during precisely the period when AI deployment is accelerating fastest. The American Federation of Teachers has urged Congress to reject the bill outright, citing concerns about worker protections.

The bill's supporters frame it differently: without a unified federal floor, they argue, companies face an increasingly unworkable maze of conflicting state requirements — more than 40 states have introduced AI-related legislation in the past 18 months. A consistent national standard, the argument goes, is better for both innovation and safety than 50 competing regulatory regimes.

Context and Timing

The draft arrives days after President Trump signed a separate executive order on June 2 promoting AI innovation and security, focused primarily on cybersecurity applications. That order did not address state preemption. The Obernolte-Trahan bill goes considerably further and represents the most detailed bipartisan legislative proposal on AI governance to reach this stage in Congress.

The document is explicitly labeled a "discussion draft" — a signal that the sponsors are soliciting feedback from industry, civil society, and other lawmakers before formally introducing it. No floor vote is scheduled, and the bill will almost certainly be revised substantially before any formal action.

Still, the release marks a shift. After years of largely informal hearings and high-level statements, Congress now has a concrete bipartisan text on the table — one specific enough to actually negotiate over.