AI Regulation Tracker / New law
Illinois Becomes First State to Mandate Independent Third-Party Safety Audits of Frontier AI Developers
Regulatory summary: The Artificial Intelligence Safety Measures Act, enacted as Illinois Public Act 104-0538 (Senate Bill 315) and signed by Governor JB Pritzker on July 6, 2026, is a binding state law that regulates large frontier AI developers, meaning companies training very-high-compute models that also generate more than $500 million in prior-year revenue. Both prongs of that test must be met, so the Act binds a short list of frontier labs and imposes no direct duty on the professionals and enterprises that merely use their models.
Governor JB Pritzker signed Senate Bill 315, the Artificial Intelligence Safety Measures Act, on July 6, 2026. The law requires the largest frontier AI developers to publish catastrophic-risk frameworks, file transparency reports, report safety incidents within 72 hours, and submit to annual conflict-free third-party audits. It takes effect January 1, 2027, with the framework and audit duties phasing in January 1, 2028.
Primary source: SB 315 enrolled text, Public Act 104-0538 (PDF)
Key takeaways
- Illinois is the first US state to require independent third-party safety audits of covered frontier AI developers.
- The scope test has two parts and both must be met: more than $500 million in prior-year revenue, and training a frontier model using more than 10 to the 26th floating-point operations. Firms that only use AI tools are not directly regulated.
- Catastrophic risk means more than 50 deaths or serious injuries, or more than $1 billion in property damage or loss, from a single incident. Some early summaries reported that figure as $1 million; the enrolled text says $1 billion.
- The Attorney General enforces the Act exclusively, with civil penalties up to $1 million for a first violation and up to $3 million for each subsequent violation. There is no private right of action.
- Signed July 6, 2026 as Public Act 104-0538. In force January 1, 2027; framework and audit duties phase in January 1, 2028.
Who the law actually binds
The Act reaches a defined category it calls a "large frontier developer," and two conditions both have to be met. First, the developer, together with its affiliates, had "annual gross revenues in excess of $500,000,000 in the preceding calendar year." Second, it trains a "frontier model," which the statute defines as a foundation model "trained using a quantity of computing power greater than 10 to the 26th integer or floating-point operations."
That combination is deliberate. It describes a short list of well-capitalized labs, not a law firm running a document review tool, not an advisory practice using a research assistant, and not a brokerage generating listing copy. If your business sits on the buying side of AI, you are outside the definition and the Act puts no direct obligation on you.
What the covered developers now owe
Beginning January 1, 2028, a large frontier developer must "create, implement, publish" and annually review a frontier AI framework, defined as "documented technical and organizational protocols to manage, assess, and mitigate catastrophic risks." From the same date, the developer must "annually retain a third party to perform an independent audit of compliance," and the auditor cannot hold a financial interest or be paid on the basis of the result.
The statute also builds a reporting duty. A developer must report a critical safety incident "to the Agency and the Attorney General within 72 hours" of forming a reasonable belief that one occurred. If an incident "poses an imminent risk of death or serious physical injury," the disclosure window drops to 24 hours to a law enforcement or public safety authority. Covered developers also owe confidential internal reporting channels and protection for employees who raise safety concerns.
How the Act defines catastrophic risk
The phrase doing the heavy lifting is "catastrophic risk," and the statute sets it at a scale that clarifies the law's intent. It means a foreseeable and material risk that a model "will materially contribute to the death of, or serious injury to, more than 50 people or more than $1,000,000,000 in damage to, or loss of, property arising from a single incident."
The enforcement design
Enforcement runs through the Illinois Attorney General, exclusively. Civil penalties reach up to $1,000,000 for a first violation and up to $3,000,000 for each subsequent violation. The Act closes the door on plaintiffs: "Nothing in this Act shall be construed to establish a private right of action associated with violations of this Act." A client cannot sue a developer under this statute, and neither can anyone else. Only the state can bring the case.
What this means if you deploy AI professionally
Nothing in the Act creates a new duty for the lawyer, advisor, agent, or clinician using AI. The obligations sit entirely with the model builder. Two steps still belong to the deployer, and both are narrower than a compliance program. Read the one paragraph in your AI vendor contract that assigns responsibility for model failures, and keep your own human review step in place regardless of what any state requires of the developer.
Buyers do gain something concrete from the statute. The models you depend on will carry more published documentation about their own limits, which is useful evidence when you need to explain a tool to a client, a regulator, or a court. Enterprise governance teams should fold those developer disclosures into vendor due diligence and AI-governance evidence now, rather than waiting for the 2028 phase-in.
Illinois in the wider state picture
A law that does not bind you can still change the ground you stand on. Illinois now joins California, which put its own frontier transparency law, SB 53, into force earlier this year. When Illinois takes effect it sits alongside New York and California as a bloc that, by the Governor's own framing, covers a large share of the American market and pushes toward a de facto national standard set by states rather than Washington.
The direction of travel is unmistakable. Responsibility for AI output is being assigned deliberately, developer duties in one place and professional duties in another. The Illinois Act puts a heavy obligation on the builder. It leaves the duty to check the work exactly where it already sat, with you.
Frequently Asked Questions
Does the Illinois AI Safety Measures Act apply to my business if I use AI tools?
Almost certainly not. The Act binds "large frontier developers" with more than $500 million in prior-year revenue that also train models above the compute threshold. A professional or firm that uses AI tools falls outside that definition and takes on no new duty under the statute.
Is the Illinois AI Safety Measures Act in force now?
It is enacted law. Governor Pritzker signed Senate Bill 315 on July 6, 2026, and it became Public Act 104-0538. The Act takes effect January 1, 2027, and the catastrophic-risk framework and annual audit duties phase in January 1, 2028.
Which companies does it regulate?
Large frontier developers. Both prongs of the test must be met: annual gross revenues above $500,000,000 in the preceding calendar year, and training a frontier model with more than 10 to the 26th floating-point operations of compute. Smaller labs and ordinary enterprise deployers are not the directly regulated parties.
What is the third-party audit requirement?
Covered developers must undergo an annual independent safety audit conducted by experts without financial conflicts of interest, who cannot be paid on the basis of the result. It is the first such state-law mandate in the United States and phases in on January 1, 2028.
How fast do developers have to report a safety incident?
Within 72 hours. A covered developer must report a critical safety incident to the Agency and the Attorney General within 72 hours of forming a reasonable belief that one occurred. If the incident poses an imminent risk of death or serious physical injury, disclosure to a law enforcement or public safety authority is due within 24 hours.
How is "catastrophic risk" defined?
As a foreseeable, material risk that a model materially contributes to the death or serious injury of more than 50 people, or more than $1 billion in damage to or loss of property, arising from a single incident. Some early summaries reported that figure as $1 million; the enrolled text says $1 billion.
What are the penalties, and can individuals sue?
The Illinois Attorney General has exclusive authority to bring an action. Civil penalties reach up to $1 million for a first violation and up to $3 million for each subsequent violation. The Act states that nothing in it creates a private right of action, so there is no individual lawsuit path.
Internal links
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.