Illinois has become the first state to require annual independent audits of the largest frontier AI developers. Governor JB Pritzker signed SB 315, the Artificial Intelligence Safety Measures Act, on July 6, 2026, and its core obligations for the biggest companies take effect on January 1, 2028. The law is now one of three state frontier-safety statutes, alongside California and New York, and it goes further than either on outside verification.

Who the Law Covers

SB 315 is built around the idea of a frontier model. According to law firm analyses of the statute, a frontier model is one trained with more than 10^26 floating-point operations. Anyone who trains such a model counts as a frontier developer. A subset of those companies, called large frontier developers, have annual gross revenues above $500 million in the preceding calendar year.

The reach extends beyond Illinois. The law applies to any developer whose models are accessible to users in the state, regardless of where the company is based. That broad scope means national and international AI companies with Illinois users will likely have to comply.

What Developers Must Do

The obligations are layered by company size. Every frontier developer must publish transparency reports before deploying a new model or substantially modifying one. Those reports cover contact information, release dates, supported languages and modalities, intended uses and generally applicable restrictions.

Large frontier developers face more. Starting January 1, 2028, they must publish, implement and update each year a safety framework covering catastrophic risk assessment and mitigation, cybersecurity and the response to critical safety incidents. They must also undergo an annual independent audit of their model risks and mitigations. The auditor and the developer may not have a financial interest in each other. A high-level summary of the audit must be posted publicly within 30 days, and the full report must be provided to the state agency and attorney general on request.

Incident reporting is faster than most people expect. Developers must report critical safety incidents to the state within 72 hours of learning facts that reasonably indicate an incident occurred. That window shortens to 24 hours when an incident poses an imminent risk of death or serious physical injury.

The law also protects employees. It bars retaliation against employees, contractors and affiliates who report safety concerns or legal violations, and it requires companies to offer internal anonymous reporting channels.

How It Will Be Enforced

The attorney general enforces the statute and may seek civil penalties. Analyses of the bill note that there is no private right of action for most provisions, meaning individuals cannot sue developers directly for violations, though the whistleblower protections are an exception.

Some terms also remain open. Commentators note that "substantially" modified models are not defined, which leaves unclear when a developer must file a new transparency report. Practitioners will likely look to the state agency's rules and early enforcement actions to fill in those details.

Why It Matters

The audit requirement is the headline. Most AI safety commitments so far have been self-published, with companies describing their own testing and review. Requiring a third party with no financial stake in the developer creates an outside check that does not depend on the company's own account. Pritzker framed the bill as a response to public demand, saying people "want protections from the risks of AI," and he presented it as a bipartisan effort.

The law also lands during a national fight over who should write AI rules, including federal efforts to limit state regulation. Illinois' approach, with its clear thresholds and fixed compliance dates, gives developers a concrete target and gives other states a template to compare against. It also raises the question of whether the largest companies will meet the same standard across several states at once.

For smaller developers, the law draws a line. Frontier model developers below the revenue threshold still face transparency reports, but they avoid the audit and framework mandates. That split means the heaviest obligations fall on the companies with the most resources to absorb them.

What to Watch

The next milestones are regulatory rather than legislative. Watch for the state agency's implementation guidance, the first published audit summaries and any incident reports that reveal how the 72-hour rule works in practice. Those signals will show whether SB 315 is a model for other states or a hard-to-enforce standard that becomes a benchmark for more ambitious rules later.

Sources