Six months ago, President Trump signed an executive order telling states to back off AI regulation. The message was clear: federal government only, one national standard, no patchwork of state rules cluttering the AI landscape. The race with China for AI superiority demanded it, the White House said.

States are not listening.

More AI-related bills have been introduced this year than last, according to Justine Gluck of the Future of Privacy Forum — and not just in Democratic-led states. Republican legislators are co-sponsoring AI accountability bills alongside Democrats. The executive order threatened to restrict federal funding to states with AI laws deemed “more than minimally burdensome.” No state has been taken to court over their AI legislation. No funding has been cut. The teeth of the order, such as they were, have not bitten.

What States Are Actually Doing

The legislation passing isn’t the sweeping AI safety bills that governors like Gavin Newsom vetoed in California. Those were too broad, too burdensome on industry, too eager to impose liability on developers for things that were hard to define. What’s replacing them is narrower, more targeted, and arguably more likely to survive legal challenge.

Illinois is the most striking case. A bill on Governor JB Pritzker’s desk — which passed with near-unanimous Republican support — piggybacks on California’s and New York’s model AI accountability laws. It requires developers of large advanced AI models to create protocols to prevent catastrophic outcomes: biological weapons attacks, large-scale infrastructure hacks, that sort of thing. But Illinois added a new twist: independent auditor requirements. Developers must have a third party review whether they’re actually following their own policies. That’s accountability with some teeth.

When asked about Trump’s opposition, Democratic state Senator Mary Edly-Allen responded: “I don’t know if you’ve met Illinois, but we’re pretty independent.” That kind of defiance from a Democrat in a swing state is one thing. What matters more is that Republican legislators went along.

Companion chatbots are a growing legislative focus across both red and blue states. Colorado, Connecticut, Idaho, Iowa, Nebraska, and Oregon have all passed laws restricting how AI can interact with people, especially children. The rules tend to require disclosure — tell users they’re talking to AI, not a human — and impose restrictions on chatbots that sustain ongoing relationships with minors. Connecticut’s new provisions prohibit companion chatbots from interacting with under-18s unless they’re programmed against encouraging self-destructive behaviour.

Colorado took a different angle: transparency in consequential decisions. Companies deploying AI in hiring, education, housing, or banking must now disclose when AI is being used to influence a decision about you. It’s a right-to-know rule for algorithmic judgment. The state’s 2024 attempt at stronger anti-discrimination provisions was watered down under pressure from Governor Jared Polis, but the disclosure framework survived.

The Federal Vacuum Problem

What’s striking about this wave is not just its breadth but its bipartisanship. The federal government has a policy framework that wants Congress to preempt state AI laws and pass national legislation on children, IP, and free speech. Congressional Republicans and Democrats are both criticizing the latest bipartisan draft proposal. Nothing has passed. The executive order exists but isn’t being enforced.

The result is a regulatory vacuum that states are filling because the alternative is no oversight at all. And unlike the failed California safety bills that alarmed industry, the current generation of state laws tends to be procedural — disclosure requirements, audit requirements, opt-out rights — rather than trying to impose liability for AI behaviour that nobody quite knows how to define.

Florida tried to advance Republican Governor Ron DeSantis’s AI “Bill of Rights” — which included chatbot restrictions and parental controls — and the state House refused to pass it. Not because the ideas were controversial, but because House Speaker Daniel Perez said the federal government should handle it. Florida’s legislature, at least, got the memo. Illinois’s didn’t.

Trump’s AI executive order told states to step back. They’re stepping forward. The federal framework aimed to be the single national voice on AI. It sounds a lot like the dog that caught the car.