# Illinois Becomes the First State to Mandate Annual AI Safety Audits > Source: > Published: 2026-07-07 05:17:17+00:00 *Illinois just told the biggest AI labs in the country that safety promises aren't enough anymore. Starting in 2028, someone else has to check the work, every year.* Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act, known as SB 315, into law on July 6, becoming the first state in the country to force independent, third-party audits of frontier AI systems. The bill sailed through the General Assembly with bipartisan support. Only five Republican senators voted against it in the Senate, and it cleared the House unanimously. The law applies to developers with more than $500 million in annual revenue whose models are trained using enormous amounts of computing power, a threshold that lands squarely on OpenAI, Anthropic and a handful of others. Those companies now have to publish a plan for addressing catastrophic risks, update it annually, and disclose how they detect and respond to what the law calls "critical safety incidents." If one of those incidents happens, the company has 72 hours to report it to the state. The audit requirement is what actually separates SB 315 from everything that came before it. New York passed its own frontier AI law, the RAISE Act, which requires a single independent audit at the point a developer becomes big enough to qualify. Illinois goes further. Covered developers must hire an outside auditor to check their compliance every year starting in 2028, and that auditor has to sign a certification. Within 30 days of receiving the report, the company has to post a redacted summary on its own website and send copies to the Illinois Attorney General and the state's Emergency Management Agency. What's notable is that the companies being regulated didn't fight it. OpenAI spokesperson Jamie Radice called SB 315 "a thoughtful framework for frontier AI safety." Anthropic went further, saying the law "takes the safety practices leading labs already follow voluntarily, publishing a safety framework, transparent reporting, protecting whistleblowers, and helps establish a baseline that every leading AI developer is expected to meet." That's a company endorsing a law that will require someone else to audit its own safety claims. Frankly, that tells you the labs would rather have one clear state standard than fifty different ones creeping in behind it. That preference runs straight into a collision the industry hasn't resolved yet. In December 2025, President Trump signed an executive order establishing a national AI policy framework and creating a Department of Justice AI Litigation Task Force with one job: challenge state AI laws that conflict with federal priorities. The order frames a 50-state patchwork as a threat to American AI dominance and directs federal agencies to identify state requirements worth fighting in court. An earlier attempt to write a ten-year moratorium on state AI laws into the One Big Beautiful Bill Act passed the House but died in the Senate, where bipartisan objections centered on states losing authority over consumer protection. So SB 315 now exists in the same window as a White House task force built to sue states over exactly this kind of law. The executive order carves out exemptions for child safety and a few other categories, but a general frontier-model audit mandate isn't obviously one of them. Colorado's AI Act and California's Transparency in Frontier Artificial Intelligence Act are already in effect and, for now, unchallenged. SB 315 doesn't take effect until January 1, 2027, which gives both sides a year to sort out whether this becomes a courtroom fight before a single audit ever gets filed. For Anthropic and OpenAI, the practical math is straightforward even if the politics aren't. Building or hiring an audit team, documenting a catastrophic risk framework, and standing up a 72-hour incident reporting pipeline all cost money, but it's a rounding error next to what either company spends on compute in a single quarter. The real cost is precedent. If Illinois's model holds, other states will copy it, and the labs that just praised SB 315 as a reasonable baseline will have to decide whether reasonable in Springfield still looks reasonable in Sacramento, Albany and wherever comes next. **Also read:** [An AI Agent Ran a Ransomware Attack Almost Entirely by Itself](https://startupfortune.com/an-ai-agent-ran-a-ransomware-attack-almost-entirely-by-itself/) • [US Investors Can Finally Buy Into SK Hynix Ahead of Its Nasdaq Debut](https://startupfortune.com/us-investors-can-finally-buy-into-sk-hynix-ahead-of-its-nasdaq-debut/) • [Syntiant files to go public just as Wall Street starts doubting the AI chip boom](https://startupfortune.com/syntiant-files-to-go-public-just-as-wall-street-starts-doubting-the-ai-chip-boom/)