SB 315 requires annual third party safety audits of large AI models. The test is whether covered developers accept them or simply stay under the $500 million revenue threshold.
Illinois Gov. JB Pritzker signed the Artificial Intelligence Safety Measures Act on Monday, putting the state on the front line of U.S. AI accountability with what supporters call a first-in-the-nation mandate: annual, independent third-party safety audits of the largest AI models.
The law, Senate Bill 315, targets the developers whose models already operate at a scale where a single failure can have national consequences. Coverage applies to systems that generate more than $500 million in annual revenue and are trained with a compute footprint Illinois defines as "massive." For those companies, the bill requires a yearly audit by a qualified, independent reviewer, separate from any internal safety team, that files its findings with the state.
The audit has to do real work. According to the enrolled bill text and the Capitol News Illinois report on the signing, the review has to assess whether a covered model materially helps with the highest-stakes misuse cases: chemical, biological, nuclear, or critical-infrastructure cyberattacks. The law also requires developers to file incident reports when their models cause specified large-scale harms. Both pieces are designed to push evidence of model behavior out of internal safety memos and into a public record the state can examine.
The mechanism is harder to fake than a self-certification regime because the auditor, not the developer, owns the conclusion. But the law has an obvious escape valve, and it sits in the same sentence as the audit requirement. Coverage is triggered by revenue and compute, two metrics a frontier-model lab can influence by where it routes sales and how it books training runs. A startup that expects to cross the threshold can structure the launch to delay it. A large incumbent can carve out the relevant products into a separate legal entity that falls just under the line. As Capitol News Illinois reported when the bill cleared the legislature, advocates treat SB 315 as a floor, not a ceiling.
Pritzker made the political economy of the bill explicit at the signing. Capitol News Illinois reported on Monday that he accused federal lawmakers of being "captive to special interests" profiting from unregulated AI, a line the syndicated regional coverage carried the same day. The Illinois Senate had passed the bill on May 21, 2026, according to the General Assembly's vote history. Pritzker's critique matters because it identifies the constituency the bill is built against: the handful of frontier developers whose Washington lobbying has, in his telling, kept Congress from passing a comparable framework.
That puts Illinois, California, and New York in the same regulatory club. SB 315 is modeled on California's SB-53 and New York's Responsible AI Safety and Education Act, both signed in late 2025, and lawmakers estimate the three states together account for roughly 40% of the U.S. AI market. The figure is a lawmaker estimate, not an audited statistic, but the practical point is that any developer who wants to sell into the largest state-level markets now faces three overlapping audit regimes with their own definitions of a "covered model." The Transparency Coalition's analysis of the law and the FMG Law write-up both flag the harmonization problem as the first major compliance question for in-house counsel.
The first round of audit reports will be the real test. If the findings produce disclosures the public and the legislature can act on, the mandate has teeth. If they settle into paperwork, the law has a different kind of problem. The Illinois General Assembly's bill status page lists the law as enacted, and the first audits will set the baseline for what "independent" actually means in Illinois.