Illinois lawmakers have put the country's strongest frontier AI audit bill on Governor JB Pritzker's desk. If he signs it, safety promises from OpenAI, Anthropic and Google DeepMind won't be enough on their own anymore.
Illinois has not just joined the state-level fight over AI safety. It has raised the price of admission. SB 315, the Artificial Intelligence Safety Measures Act, passed the Illinois House in late May after clearing the Senate, and Pritzker said publicly that he plans to sign it. According to WIRED, the bill would make Illinois the first state to require independent, third-party audits of frontier AI developers' safety practices.
The bill is aimed at the companies building the largest models, not every startup with a chatbot. It covers developers above $500 million in annual revenue whose systems are trained with very large amounts of computing power, a line that points toward OpenAI, Anthropic, Google DeepMind and Meta. Those companies would have to publish a safety framework, explain how they assess dangerous model capabilities, report critical safety incidents, and show how they follow the standards they claim to follow.
The audit requirement is the point. California's Transparency in Frontier Artificial Intelligence Act and New York's RAISE Act put real disclosure duties on large frontier AI developers, including safety frameworks, public reports and incident reporting. Illinois goes further by bringing in an outside auditor to check whether the lab is actually doing what its own framework says it does. That changes the relationship between the companies and the public. You don't have to take the lab's word for it.
As WIRED reported, Secure AI Project policy director Scott Wisor put the problem plainly: AI companies have been grading their own homework. That line works because it's true. If a frontier lab says it has tested a model for catastrophic risk, there's no way for most people to check it: not lawmakers, not customers, not the person using the product. SB 315 doesn't solve every safety problem in AI, and it doesn't pretend to. It creates a check on self-certification, which is exactly where the current system is weakest.
What's notable is that some of the companies being regulated didn't fight the bill. OpenAI's Chris Lehane told WIRED that the Illinois General Assembly had shown bipartisan leadership by advancing SB 315 and building a framework around safety, transparency, incident reporting and accountability. Anthropic said it supported the bill too, with Cesar Fernandez, its head of U.S. state and local government relations, saying the measure would help establish a baseline every leading AI developer is expected to meet.
Frankly, that support tells you something. The largest labs would rather deal with one narrow audit law than a messy patchwork of liability rules and disclosure mandates, with emergency bills arriving state by state on top. They know Congress has not passed a broad AI safety law. They also know voters are no longer willing to let the most powerful model developers set their own rules and call that governance.
Not everyone in tech is on board. WIRED reported that Chamber of Progress, a trade group whose partners include Google, Apple, Amazon and Andreessen Horowitz, urged Illinois lawmakers to oppose SB 315. Its chief executive, Adam Kovacevich, argued the bill would expose sensitive systems to untested auditors in a regime with liability but no clear standards. That concern is not imaginary. Frontier AI audits are still a young field, and auditors will need access to information companies don't want scattered around.
Still, secrecy cannot be the whole answer. If a company is building a model powerful enough to fall under SB 315, it can afford a serious audit process. OpenAI and Anthropic already spend enormous sums on compute, safety staff and policy teams. The cost of documenting a catastrophic risk framework and submitting it to outside scrutiny is not what will break them. The precedent is what matters.
That precedent now runs into Washington. President Trump signed an executive order on December 11, 2025, directing federal agencies to push toward a national AI framework and challenge state AI laws that conflict with federal priorities. Politico reported that the order empowered agencies including the Justice Department, the Federal Trade Commission and the Commerce Department to take aim at state rules. The order also reflected the administration's complaint that a 50-state patchwork could slow American AI development.
Illinois is now testing that argument in a concrete way. The bill is not a broad ban on AI, and it is not a licensing regime for every model launch. It asks the biggest developers to publish safety plans, report serious incidents and let someone outside the company check compliance. If the White House wants to call that an obstacle to American AI leadership, it should have to say so in plain language.
The timing matters too. Earlier attempts in Congress to block state AI regulation, including a proposed ten-year moratorium tied to the One Big Beautiful Bill Act, ran into bipartisan resistance in the Senate. State lawmakers heard the message. California moved first on frontier AI transparency. New York followed with the RAISE Act. Illinois is now trying to make independent auditing part of the baseline.
For readers, the practical takeaway is simple. If SB 315 becomes law, the fight over AI safety moves from voluntary pledges to verifiable process. That won't satisfy people who want hard limits on model development, and it won't satisfy companies that want one light federal rule. But it gives the public something better than a press release. It gives you a nameable person outside the lab whose job is to check the work.
Also read: An AI Agent Ran a Ransomware Attack Almost Entirely by Itself • 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
