Illinois just became the first US state to force AI developers to prove their safety claims instead of simply publishing them. On July 6, Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act (SB 315), and now OpenAI, Anthropic, Google, Meta and xAI all have to hire outside auditors every year to check the risk profile of their most powerful models.
Until now, companies mostly graded their own homework. California and New York already required a public risk-management plan, but nobody checked whether firms actually followed it. Illinois raised the bar: the law covers models trained with more than 10^26 floating-point operations, built by developers with over $500 million in annual revenue. For those companies, an independent audit is no longer optional, and a summary of the results has to go up on the company's website within 30 days.
There's teeth behind the paperwork, too. If a model starts behaving unpredictably or creates a safety risk, the company must notify state authorities within 72 hours, or 24 hours if the danger is imminent and could cause death or serious injury. Employees who blow the whistle on violations get explicit protection from retaliation. Companies that skip an audit, miss a deadline or mislead regulators face fines of up to $1 million for a first violation and $3 million for repeat offenses, enforced by the state attorney general.
The timing isn't an accident. Congress has been quietly debating a federal bill that would freeze new state AI development rules for three years, still just a draft, not law, but Illinois clearly wanted to lock in its rules before that window closes. Both OpenAI and Anthropic backed SB 315 publicly, but for different reasons: Anthropic wants states to keep passing tougher laws precisely so Congress feels pressure to write a strong federal standard rather than a watered-down one, while OpenAI would rather have a single national rulebook than fifty different versions to comply with.
The law takes effect January 1, 2027, giving companies about a year and a half to line up auditors and build reporting pipelines. If Illinois survives the inevitable industry pushback, expect other statehouses to copy its playbook — mandatory AI audits could become an industry norm well before Congress passes anything at all.
