Days after President Trump declined at the last minute to sign an executive order setting up even a voluntary AI safety-testing framework, the Illinois legislature did the opposite. It passed SB 315, a bill that would make Illinois the first state in the country to require independent, third-party audits of the safety practices of frontier AI developers. The vote was not close: 110 to 0 in the House, 52 to 5 in the Senate. Governor JB Pritzker has signalled he will sign it, after which the Legislature has 30 days to send it over.
The bill borrows its core from laws already on the books in California and New York, requiring companies such as OpenAI and Anthropic to publish and annually update plans for managing severe or catastrophic risks from their models. What makes Illinois first is the audit mandate: outside reviewers, not the companies themselves, would assess whether those safety practices are real rather than rhetorical. SB 315 also creates whistleblower protections for employees and civil penalties for violations, and it would take effect on 1 January.
The striking part is who lined up where. OpenAI and Anthropic, the two largest American AI labs and the most obvious targets of the law, publicly supported it. A trade association representing other AI companies opposed it; Google, xAI and Meta did not comment. Anthropic's head of state and local government relations, Cesar Fernandez, framed the bill as taking "the safety practices leading labs already follow voluntarily" and turning them into "a baseline that every leading AI developer is expected to meet." An OpenAI spokesperson praised the legislature's "bipartisan leadership." The companies best able to absorb a new compliance regime were among its loudest backers.
That alignment is less surprising than it first appears. A mandatory floor codifies what the leaders already do voluntarily, while raising the cost of entry for smaller rivals who do not yet carry safety teams and audit budgets. It also hands the labs a seat at the table in shaping rules they expect to face eventually, and a more predictable regime than the alternatives. The resistance, tellingly, came from the wider field and from Washington, where the White House has argued that a patchwork of differing state laws would hamstring the American AI industry.
The bill's own sponsor half-agrees. "The states shouldn't be doing this," Representative Daniel Didech told NBC News; "the best way to regulate these types of catastrophic risks would be a federal approach." But Congress has not acted, and so, he said, states "have had no choice but to step in." That is the thread connecting Illinois to the wave of repeals elsewhere. With no federal rulebook, the rules are being written state by state, unevenly and in real time. The recent story has been of governments unwinding their own AI laws; Illinois is the reminder that the retreat is not universal, and that some of the companies being regulated would now rather help hold the pen than knock it away.