A bipartisan group of House lawmakers on Thursday released a 269-page discussion draft that would bar states from passing new laws regulating the development of artificial intelligence for three years, while creating the first comprehensive federal audit regime for frontier A.I. labs. The Great American Artificial Intelligence Act of 2026, unveiled by Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.), with co-sponsors Suhas Subramanyam (D-Va.), Scott Franklin (R-Fla.), Scott Peters (D-Calif.) and Erin Houchin (R-Ind.), arrives two days after President Trump signed an executive order establishing voluntary 30-day federal reviews of new frontier models. The sequencing isn’t coincidence; it’s choreography.

The bill imposes new oversight obligations on developers with more than $500 million in annual gross revenue, the bright line for what the text considers a frontier lab. It authorizes $100 million annually from fiscal 2027 through 2029 for the Commerce Department’s Center for AI Standards and Innovation, the office the Biden-era U.S. AI Safety Institute was rebranded into last year. It codifies the National Artificial Intelligence Research Resource. It extends the Cybersecurity Information Sharing Act of 2015 through fiscal 2035. And it defines “catastrophic risk” as “a foreseeable and material risk of death or injury.”

The preemption mechanic is where the politics live. According to an accompanying document released by Trahan’s office, California’s AB 2013, which requires model developers to publish summaries of training data, would be preempted; portions of California’s SB 942 watermarking law would be preempted; and frontier safety statutes in California, New York and Illinois would be “federalized.” The bill text itself draws no distinction between the two. State attorneys general retain enforcement authority, a concession that the drafters are clearly betting will soften the blow.

It hasn’t. “This bill takes the current floor on state A.I. legislation and turns it into a federal ceiling,” said Brad Carson, president of Americans for Responsible Innovation and a former Democratic representative from Oklahoma, who called the preemption provision a “generational mistake.” Brendan Steinhauser, chief executive of the Alliance for Secure AI, said the federal framework “does not justify preempting states’ ability to pass their own A.I. safeguards.” Industry’s reception was warmer: Jason Oxman of the Information Technology Industry Council praised “key elements that will accelerate American A.I. leadership.”

The lawmakers said they intend to “solicit feedback from stakeholders, experts, and the public before the bill is formally introduced.” That’s the public posture. The structural posture is harder to miss. The 2026 draft borrows the architecture of the 1996 Telecommunications Act, which similarly traded federal preemption for a regulatory floor and produced two decades of litigation over what counted as state authority. The bet, then as now, is that uniformity is worth more than state-level experimentation. California’s legislature, having spent two years building the very statutes this bill would erase, is unlikely to agree.

Sources