IMPLICATOR
Illinois Gov. JB Pritzker signed SB 315 on July 6, making Illinois the first state to require the largest AI developers to undergo independent safety audits every year. Anthropic wants successive state bills to add stronger duties, Cesar Fernandez, the company's head of U.S. state and local government relations, told POLITICO. OpenAI also backed Illinois, but its reverse-federalism plan asks states to converge on a common baseline for a national framework.
Fernandez said Anthropic also wants federal rules, but that a government response "can't wait for action in Washington." California's 2025 law required large developers to publish safety plans and report safety incidents. New York's RAISE Act added an independent audit when a developer first qualified. Illinois made the audit annual. "Transparency and self-reporting, we don't believe are sufficient anymore," Fernandez added.
Key Takeaways
- Illinois now requires annual independent safety audits from frontier AI developers earning more than $500 million.
- Anthropic wants successive state bills to add stronger duties, while OpenAI favors a common baseline.
- The laws govern developer safety practices and do not direct buyers toward Claude or ChatGPT.
- Groups aligned with the rival labs spent more than $23 million in one New York primary.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
The Illinois law applies to frontier developers with annual gross revenue above $500 million, including OpenAI, Anthropic, Google, Meta and xAI, according to Information Security Media Group. Covered firms must publish and annually update a framework assessing catastrophic risk and cybersecurity, then file a transparency report before deploying a new or substantially modified frontier model. Critical safety incidents carry a 72-hour reporting deadline, shortened to 24 hours when there is an imminent threat of death or serious physical injury. The attorney general may seek up to $1 million for a first violation and $3 million for later violations.
OpenAI says states should converge on a common standard. Its May post described California, New York and Illinois as a "de facto national standard" and presented Illinois' independent audits as an extension of that baseline. The company assigns prerelease testing to federal institutions and postdeployment reporting and accountability to states.
Anthropic's June policy framework asks Congress to preserve state authority until it passes rules at least as strong as the company's proposal. The document calls for public test summaries, regular independent reviews, security programs covering model weights and training infrastructure, and government authority to block or deter dangerous deployments. Fernandez cited Anthropic's internal tests, which the company said showed Claude Mythos exploiting security flaws in every major computer operating system.
Get Implicator.ai in your inbox
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
California, New York and Illinois govern safety plans, audits and incident reporting; none directs buyers toward Claude, ChatGPT or another model. David Sacks, a venture capitalist and former Trump AI and crypto czar, accused Anthropic on X of "running a sophisticated regulatory capture strategy based on fear-mongering." Fernandez told NBC News that Illinois' requirements mirror safety-testing protocols leading developers already perform voluntarily.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
Speaking to The Hill, Graham Dufault, general counsel for the Association for Competitive Technology, said differences among state disclosure rules "will add up" and could pull smaller companies into their scope. Suzanne Borders, CEO and founder of BadVR, described compliance in the same report as a "legal exposure headache" for her company; she was unsure whether she was answering the questions properly or would be sued.
NPR reported that groups associated with OpenAI investors spent against New York Assembly member Alex Bores, a co-sponsor of the state's AI law, while Anthropic-backed groups spent in his favor; combined spending exceeded $23 million. Bores lost to Micah Lasher, another sponsor of the law.
In late June, Anthropic endorsed language in a Massachusetts economic development bond bill that would require leading AI companies to hire independent evaluators for catastrophic-risk assessments, including whether a model could assist in developing bioweapons. The proposal would give the state attorney general authority to enforce the mandate. OpenAI spokesperson Liz Bourgeois said the company supported the legislature's focus on safeguards but was still reviewing the proposal. Massachusetts lawmakers are still developing the bond bill, and OpenAI has not announced a final position on its evaluator mandate.
Frequently Asked Questions
What does Illinois SB 315 require?
Frontier AI developers with annual gross revenue above $500 million must publish and update a safety framework, undergo annual independent audits, file reports before major model deployments and report critical incidents within 72 hours, or 24 hours for imminent threats to life or safety.
How do Anthropic and OpenAI differ on state AI laws?
Anthropic wants successive state bills to impose stronger safety duties. OpenAI also supported Illinois, but it wants states to converge on a common baseline that can inform one federal framework.
Do these laws favor Claude or ChatGPT?
The cited provisions regulate developer audits, safety plans and incident reporting. They do not require governments or companies to buy Claude, ChatGPT or any other model.
Why are smaller technology companies concerned?
Graham Dufault of the Association for Competitive Technology said differences among state disclosure rules can add up and pull smaller companies into their scope. BadVR CEO Suzanne Borders called the compliance burden a legal exposure headache.
What is Massachusetts considering next?
Draft language in an economic development bond bill would require leading AI companies to hire independent evaluators for catastrophic-risk assessments and would let the state attorney general enforce the mandate. OpenAI says it is still reviewing the proposal.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
