AI Safety Measures Act (frontier model audits)
Illinois SB 315 — Artificial Intelligence Safety Measures Act
Illinois · IL SB315 (104th General Assembly, 2025-2026)
- Status
- Enacted (not yet in effect)
- Takes effect
- January 1, 2027
- Applies to
- Large "frontier" AI developers meeting the Act's covered-developer thresholds (the largest model developers).
- Enforced by
- Illinois Emergency Management Agency and Office of Homeland Security; Illinois Attorney General
Illinois is the first state to require independent third-party safety audits of the largest 'frontier' AI developers (companies like OpenAI, Anthropic, and Google DeepMind). Covered developers — those with >$500M annual gross revenue whose models meet defined compute thresholds — must publish and annually update a frontier AI safety framework addressing catastrophic risks (defined as incidents threatening 50+ deaths, serious injuries, or $1B+ in damages), file transparency reports before deploying new or substantially modified models, report critical safety incidents within 72 hours (24 hours for imminent harm), and protect whistleblowers. Enforced by the Illinois Emergency Management Agency and Office of Homeland Security with the Attorney General; civil penalties; no private right of action. Law takes effect January 1, 2027; audit requirements operative January 1, 2028.
Who is protected: Illinois public and downstream users exposed to catastrophic risks from frontier AI models; covered-employee whistleblowers at frontier developers.
Legal detail
Creates the Artificial Intelligence Safety Measures Act (P.A. 104-___). Signed July 6, 2026. Law effective January 1, 2027; audit/transparency-report provisions operative January 1, 2028. Covered entity: 'large frontier developer' (annual gross revenue >$500M; AI system trained above defined compute thresholds — targeting OpenAI, Anthropic, Google, Meta, xAI). Key requirements: (1) publish and annually update Frontier AI Framework covering catastrophic risk, cybersecurity, governance, third-party evaluation, and internal-use risk; (2) pre-deployment transparency reports and catastrophic-risk assessment summaries; (3) annual independent third-party audit with access/reporting/retention/publication requirements; (4) report 'critical safety incidents' within 72 hours (24 hours for imminent death/serious injury risk); (5) whistleblower protections and internal reporting channels for covered employees. Amends Freedom of Information Act and Whistleblower Act. Enforcement: Illinois Emergency Management Agency and Office of Homeland Security in consultation with AG. Passed House 110-0; Senate 52-5 (2026-05-29).
Key facts
| Jurisdiction | Illinois |
|---|---|
| Level | State |
| Strength | Moderate |
| Enacted | July 6, 2026 |
| Citation | IL SB315 (104th General Assembly, 2025-2026) |
| Private right of action | No, agency enforcement only |
| Penalties | Civil penalties for violations as set by the Act. |
| Topics | automated decision-making · AI disclosure and transparency · government use of AI |
| Bill text | Illinois General Assembly — SB0315 Bill Status (104th GA) |
More AI rules in Illinois
- Illinois Children's Social Media Safety Act — algorithmic recommendations ban for minors · Enacted (not yet in effect)
- Illinois SB 3114 — Transparency in Downcoding Act · Enacted (not yet in effect)
- Algorithmic rent price-fixing ban · Expired
- Illinois HB 3851 — deepfake cyberbullying in schools (2026) · In effect
- Illinois data center tax-incentive pause (DCEO, July 2026) · In effect
- IL AV local-preemption (2017) · In effect
Related automated decision-making rules elsewhere
- CA EO N-6-26 (AI Workforce) · In effect
- Arkansas public entity AI/ADS policy mandate (Act 848, 2025) · In effect
- Connecticut algorithmic rent ban — HB 8002 (eff. Jan 1, 2026) · In effect
- San Diego algorithmic rent price-fixing ban (Ord. O-21955, May 2025) · In effect
- Minneapolis algorithmic rent ban (Ord. 2025-010, eff. Mar 1, 2026) · In effect
- Seattle algorithmic rent-fixing ban (Ord. 127241 / SMC 7.34, July 2025) · In effect
See something wrong or out of date? Submit a correction. Every entry must carry a verifiable source.