Illinois House of Representatives 在 2026-05-27 通过 SB 315，要求 OpenAI、Anthropic、Google DeepMind 等 frontier AI labs 接受第三方稽核，以确认其安全实务符合自订标准；若由州长 JB Pritzker 签署成法，将成为美国最强的 AI 安全法规之一。相较于 California 与 New York 主要要求揭露 model guardrails 与安全事故报告，Illinois 进一步要求独立稽核人员验证合规，补上「公司自己考自己的试」的缺口。
SB 315 预期可由 Deloitte、EY、KPMG、PwC 等 Big Four 会计稽核公司执行，也可能由 AI Evaluator Forum 的成员如 METR、Transluce、Averi 参与。支持者指出，这项法案提供对 frontier AI 的基准线，并可能作为未来 federal AI legislation 的测试场；Daniel Didech 强调，州立法能提高国会之后通过全国性法规的机率。
在政治与产业角力上，OpenAI 曾对一项较早的 Illinois 法案表态支持，但之后又称先前的全面支持是 oversights；Anthropic 则表示自己是第一个支持 SB 315 的 AI lab。相对地，Chamber of Progress 反对该案，认为它会迫使企业向未经验证的稽核者暴露敏感系统。背景上，Trump 第二任期已撤销多项 Biden 时期与州层级 AI 规范，理由是避免「patchwork」监管；上周他也取消一项原定签署的行政命令，表示不想削弱美国对 China 的竞争优势。
On May 27, 2026, the Illinois House of Representatives passed SB 315, a bill that would require frontier AI labs such as OpenAI, Anthropic, and Google DeepMind to undergo third-party audits of their safety practices. If Governor JB Pritzker signs it, the law would become one of the strongest AI safety regimes in the United States, going beyond California and New York by mandating independent verification that companies are actually following their own safety standards.
The bill is expected to be implemented by major accounting and auditing firms like Deloitte, EY, KPMG, and PwC, though members of the AI Evaluator Forum, including METR, Transluce, and Averi, could also be used. Supporters say SB 315 creates a baseline for frontier AI safety and could serve as a testing ground for future federal policy, with Daniel Didech arguing that state laws like this make eventual national legislation more likely.
The fight over SB 315 also reflects broader industry and political conflict. OpenAI previously backed an earlier Illinois bill but later said that support had been an oversight, while Anthropic says it was the first AI lab to support SB 315. Opponents such as Chamber of Progress argue the bill would expose sensitive systems to untested auditors. At the same time, President Donald Trump’s second-term agenda has rolled back several Biden-era and state-level AI rules in favor of avoiding fragmented regulation and preserving U.S. competition with China.