Reps. Jay Obernolte, R-Calif., and Lori Trahan, D-Mass., released a 269-page discussion draft of the Great American Artificial Intelligence Act on June 4, paired with four additional bipartisan co-sponsors and a three-year preemption of state laws regulating how AI models are built. It’s the third congressional run at overriding state AI law, after a Senate moratorium attempt died 99-1.

The preemption is narrow by design and broad in effect. Per Roll Call, the bill freezes state laws “specifically regulating the development” of frontier models but leaves use-and-deployment rules alone. Trahan’s office confirms California’s AB 2013 training-data disclosure law would be wiped off the board for the duration.

In exchange, Washington gets a regime. Developers above $500 million in prior-year gross revenue would’ve to publish safety frameworks and report critical incidents. The draft codifies the Commerce Department’s Center for AI Standards and Innovation at $100 million a year through FY2027–2029, extends the Cybersecurity Information Sharing Act of 2015 to FY2035, creates criminal penalties for AI impersonation of government officials, slots AI-adoption questions into Census Bureau and BLS surveys, and stands up a DOE/NIST/NSF testbed. Reps. Scott Franklin, Suhas Subramanyam, Erin Houchin, and Scott Peters signed on.

The House Commission on AI and the Innovation Economy rejected the text within hours, saying it “does not meet the enormity of the moment” and “cannot serve as the basis for productive dialogue.” Brendan Steinhauser of the Alliance for Secure AI praised the bipartisanship but drew the same line on preemption: “A national AI standard should protect at least as much as it preempts.”

Whether the Trump administration backs any of it, per Roll Call, remains unclear. The package is a discussion draft, not introduced legislation, which is what makes the immediate Democratic-commission rebuke notable: the negotiating floor is already contested before the bill has a number.

Sources

Sources