Congress Floats a National AI Law That Would Override the States

For two years the real action in American AI regulation has happened in state capitals, where dozens of legislatures wrote their own rules while Congress watched. A new bipartisan draft would change that, and not everyone is pleased about how.

On June 4, Representatives Jay Obernolte, a California Republican, and Lori Trahan, a Massachusetts Democrat, released a 269-page discussion draft of the Great American Artificial Intelligence Act. It is the most complete federal framework for governing artificial intelligence that Congress has put on paper, and it tries to settle the question of who writes the rules: Washington or the states.

What the draft would do

The bill sets obligations for the largest model builders. Developers above $500 million in annual revenue would have to publish safety frameworks, report critical incidents, and submit to twice-yearly third-party audits, according to FedScoop. The draft also writes the Commerce Department's Center for AI Standards and Innovation into law, giving the federal government a permanent technical body to lean on.

That part has drawn relatively little fire. The fight is over a single word: preemption.

The three-year freeze

The draft would block state laws that specifically regulate the development of AI models, with the freeze lapsing after three years. As Roll Call reported, the preemption is written to spare laws that govern how AI is used or deployed, so a state could still regulate AI in hiring or medical advice. The line sits at the model itself: how it is trained, tested, and released.

Supporters argue that a single national standard beats a patchwork of fifty, and that frontier labs should not have to satisfy contradictory rules in every state where their software runs. Critics see a giveaway. Brad Carson, president of Americans for Responsible Innovation and a former Democratic congressman from Oklahoma, called the provision a "generational mistake." In his words, "This bill takes the current floor on state AI legislation and turns it into a federal ceiling, preventing state lawmakers from addressing emerging AI harms."

The tension is real. States have been the laboratory for AI rules, from Vermont and Ohio moving to limit chatbots in mental health care to broader consumer-protection bills. A three-year pause on development rules would, for that window, hand the pen to Congress and the Commerce Department alone.

A draft, not a bill

It is worth being precise about where this stands. This is a discussion draft, circulated to gather feedback from companies, researchers, and the public before anything is formally introduced. Nothing here is law yet, and the preemption language is exactly the kind of provision that gets rewritten between draft and vote.

Still, the timing matters. It lands days after the White House moved to give the government early access to frontier models for review, and it signals that federal AI policy is finally moving from speeches to statute. Whether it moves toward more oversight or less depends on which word survives markup.