The map of American AI regulation is being drawn in state capitals, not in Washington. Across the first half of 2026, legislatures from Springfield to Providence pushed through bills governing how the most powerful models get tested, how chatbots speak to children, and whether companies can quietly use AI to set prices. With Congress still debating a federal framework, the states have stopped waiting.
The sharpest signal came out of Illinois. Lawmakers sent Governor JB Pritzker SB 315, the Artificial Intelligence Safety Measures Act, which would require developers of frontier models to submit to annual independent third-party audits of their safety practices. As NBC News noted, it is the first state measure to demand outside audits rather than the self-reported assessments most companies publish today. If Pritzker signs, Illinois becomes the third state to write rules aimed specifically at frontier systems, after New York's RAISE Act and California's SB 53.
A crowded session
New York was just as busy. Having already enacted the RAISE Act, which obliges large developers to publish their safety protocols and report serious incidents to the state within 72 hours, lawmakers sent Governor Kathy Hochul seven more AI bills before the session closed on June 1. They include a children's chatbot safety measure, an AI Training Data Transparency Act requiring developers to disclose what their models learned from, a FAIR News Act covering AI-generated news content, and a ban on surveillance-based pricing.
Not every bill survived. In Colorado, Governor Jared Polis signed four AI measures but vetoed HB 1210, which would have restricted algorithmic pricing and the use of intimate personal data to draw financial inferences about people. Rhode Island moved the other way, approving H 7349, a ban on therapy chatbots, alongside bills on health-insurance AI transparency and chatbot safety features.
Chatbots draw the most fire
If one theme runs through the year, it is chatbots. By the IAPP's count, lawmakers have tracked close to 100 chatbot-related bills across 34 states, with at least a dozen already signed into law. Most center on the same handful of worries: telling users when they are talking to a machine, protecting minors, and setting crisis-response rules when someone signals they may be in danger. Georgia's SB 540, signed in May, is a representative example, requiring disclosure, limits on interactions with children, and self-harm response protocols.
Why the rush
The state-level scramble reflects a vacuum at the federal level. Congress has floated a national AI law that would override much of this state activity, but it remains a draft. The White House's recent executive order, meanwhile, leaned toward partnership with industry rather than licensing or hard mandates. Into that gap, statehouses are writing requirements that frontier developers will have to meet customer by customer, state by state.
For companies, that patchwork is the real headache. A model offered nationwide may soon face audit rules in Illinois, incident-reporting deadlines in New York, and disclosure duties in Georgia, all at once. For now, the most consequential AI laws in the country carry the signatures of governors, not the president.
Commentarii · 0