Illinois SB 315 (frontier safety framework with mandatory third-party audits)

Illinois SB 315 is a state frontier-AI safety bill modeled on California SB 53, adding a mandatory independent third-party compliance audit requirement and whistleblower protections on top of an SB 53-style transparency framework. Both OpenAI and Anthropic endorsed the bill in May 2026. The Illinois House passed it on May 27, 2026 and the full General Assembly approved it on May 28, 2026; Gov. JB Pritzker (D) signed it into law on July 6, 2026, and the statute takes effect January 1, 2027 (Source: govtech.com). It is the first U.S. state law to require independent third-party audits of frontier-AI safety practices.

Status and timeline

The Illinois House of Representatives passed SB 315 on May 27, 2026 — unanimously, per bill sponsor Rep. Daniel Didech (Source: govtech.com) — and the full General Assembly approved the bill on May 28, 2026. Gov. JB Pritzker posted on X that he will sign the bill, calling it Illinois "leading the nation in holding Big Tech accountable." The statute takes effect January 1, 2027 — three months ahead of NY RAISE Act provisions and a year ahead of Connecticut SB 5's sandbox deadline (Source: wired.com; iapp.org).

Pritzker signed SB 315 into law on July 6, 2026, making Illinois the first state to mandate annual independent third-party audits of frontier AI developers' safety practices (Source: govtech.com). The governor's office titled the enacted statute the Artificial Intelligence Safety Measures Act; the law mandates public disclosure of safety practices, incident reporting, and whistleblower protections, and carries civil penalties of up to $3 million. Coverage of the signing described Anthropic as the first AI lab to back the bill (Sources: gov-pritzker-newsroom.prezly.com; capitolnewsillinois.com). At the signing Pritzker said that with "the federal government unwilling to step in, states have a responsibility to protect our people from the dangers of AI" (Source: route-fifty.com).

Sponsor Rep. Daniel Didech told NBC News that states "have had no choice but to step in" given the pace of frontier-model development, a state-as-default-regulator framing discussed under AI Federalism, voiced as Congress's frontier-AI EO stalled the same week. Illinois joins California and New York with frontier-AI laws, deepening tension with President Trump's stated desire to preempt state-level AI regulation (draft preemption EO).

Scope and definitions

SB 315 defines "catastrophic risk" to include capacity for mass harm or more than $1 billion in cyberattack or malfunction damage. The bill's obligations fall on covered frontier developers.

Key provisions

The enacted text:

• Imposes mandatory annual third-party audits, the first such requirement among U.S. states.
• Requires governance, risk-mitigation, and cybersecurity duties on covered frontier developers.
• Requires pre-deployment reports on model capabilities, intended use, and risk profile.
• Requires critical-incident reporting within 72 hours, shortened to 24 hours where there is a risk of death or serious harm (see AI Incident Reporting) (Source: govtech.com).
• Creates whistleblower protections for employees who report safety issues, paralleling Connecticut SB 5's anonymous risk-reporting regime, but layered on top of mandatory external audits rather than replacing them.

Comparison with CA SB 53 and NY RAISE Act

SB 315 adds a mandatory independent third-party audit requirement to the SB 53-style transparency framework, the first such requirement for AI safety frameworks in U.S. state law. By contrast, the two other frontier-AI state laws rely on different enforcement structures:

• CA SB 53: AG civil enforcement; the AG cannot rulemake.
• NY RAISE Act: AG civil enforcement plus a new DFS Oversight Office that publishes annual assessments.

SB 315 layers independent third-party compliance auditors as a structural enforcement layer on top of this transparency model.

The distinction was invoked after OpenAI disclosed in July 2026 that models had escaped its isolated test environment. Secure AI Project co-founder Thomas Woodside said SB 53 and the RAISE Act require frontier developers to describe their approach to internal-use risk without specifying the content of that description, while SB 315 adds third-party audits of whether a developer actually adheres to its own plan. Woodside dated the audit obligation to January 2028, which does not match the January 1, 2027 effective date reported for the statute as a whole; whether the audit duty phases in a year after the statute takes effect has not been confirmed against the enacted text (Source: transformernews.ai). See Rogue Internal Deployment.

OpenAI and Anthropic endorsement

Both OpenAI and Anthropic publicly endorsed SB 315 in mid-May 2026 (around May 14-15) as part of what OpenAI described as "an emerging national framework alongside California and New York" (Source: openaiglobalaffairs.substack.com). This was the first known case of both major U.S. frontier labs endorsing a mandatory third-party audit requirement for their own safety frameworks. At passage, both companies continued to support the bill; Anthropic's Cesar Fernandez framed SB 315 as a baseline "every leading AI developer is expected to meet" (Source: wired.com).

The endorsements stand in tension with the same labs' federal preemption push and have not yet been tested against the bill's audit-administration design, including who administers audits and with what subpoena power. The Transformer analysis of OpenAI's position appeared on May 15, 2026 (Source: transformernews.ai).

OpenAI's June 2, 2026 federal blueprint carried the audit requirement into its proposed national framework, asking that large frontier developers "annually retain an independent third party to audit compliance with frontier safety requirements, including implementation of the developer's frontier AI framework, internal controls, and governance structures," underpinned by "a set of common standards that allow for interoperable audits across jurisdictions." The blueprint presents SB 315 alongside SB 53 and the RAISE Act as three statutes that "share common requirements," but on this particular obligation it is SB 315 alone among the three that the federal proposal tracks (Democratic Governance of Frontier AI: A blueprint for a federal framework (OpenAI, June 2026)). The same document asks Congress to preempt state laws regulating the same frontier safety risks once such a framework is enacted, which would displace the Illinois audit duty with the federal one (see AI Federalism).

OpenAI cited SB 315 again on August 7, 2026, listing its support for the statute's frontier-safety-framework audits alongside support for youth-safety audits under the Parents & Kids Safe AI Act, as evidence of a consistent position in a post setting out six principles for audit and assessment design (Making AI Audits and Assessments Work (OpenAI Global Affairs, August 2026)). That post carries the interoperability argument further than the June blueprint: it asks that "common national or international standards" define the risks and controls in scope, the evidence and criteria reviewers use, the level of assurance expected, and how assessments are conducted and reported, on the reasoning that without a shared foundation "reviewers may apply inconsistent questions and thresholds, making findings difficult to compare or act on." It also states that companies should be able to choose among qualified assessors while the system guards against "assessor shopping" — a design question the Illinois statute's audit-administration provisions have not been shown to resolve.

Opposition

The Chamber of Progress (whose partners include Google, Apple, Amazon, and Andreessen Horowitz) and the Computer & Communications Industry Association (CCIA) opposed the bill. CCIA's May 28 letter argued that "no credible or standardized ecosystem currently exists to conduct the type of independent audits envisioned" (Source: wired.com; iapp.org).

Relation to Illinois SB 3444

SB 315 is one of two Illinois AI bills moving in the same period under different frameworks. SB 315 is frontier-developer-targeted and lab-endorsed; Illinois SB 3444 — Artificial Intelligence Safety Act is a deployer-targeted bill with a liability safe harbor that OpenAI opposed. OpenAI's Caitlin Niedermeyer disavowed the liability safe harbor in SB 3444 in May 13, 2026 written testimony.

Relationships

• modeled-on: California SB 53, New York RAISE Act
• related: Illinois SB 3444 — Artificial Intelligence Safety Act (sibling Illinois bill — different framework, OpenAI disavowed safe harbor), AI Pre-Release Vetting, State-Level AI Regulation, California Effect, OpenAI, Anthropic (both endorsers)
• supports: AI Transparency, AI Safety Cases and Frameworks

Sources

• (Source: openaiglobalaffairs.substack.com) — OpenAI Global Affairs blog endorsing SB 315 + KOSA (May 13, 2026)
• (Source: transformernews.ai) — Transformer analysis, May 15, 2026
• WIRED — "Illinois Lawmakers Just Passed America's Strongest AI Safety Bill" (May 27, 2026) (Source: wired.com)
• IAPP — "Notable AI, privacy bills hit finish line in Illinois, Connecticut and New York" (May 28, 2026) (Source: iapp.org)