AI auditors get a state registry in California

Gavin Newsom signed two bills placing California's outside AI auditors under state oversight, one creating a registry for them and one setting up a state-checked credential for organizations that assess AI risk. Politico reported the signings, which Newsom and one of the bills' authors used to press Washington to act.

At a glance

• The auditors in question are outside groups that AI developers can hire to check whether their models comply with California law, and Bauer-Kahan's measure adds ethical rules governing that work.
• McNerney's bill hands the state the job of setting criteria for Independent Verification Organizations and vetting their expertise in assessing the risks that AI systems pose.
• OpenAI and Anthropic both backed the package, and OpenAI's chief global affairs officer Chris Lehane said the company will keep pressing state legislatures until Congress passes national AI safety regulation.

Auditing is the layer that determines whether the rest of AI regulation functions: rules requiring evaluation mean little without a defined set of evaluators the state recognizes. California appears to be building that layer ahead of the national regulation that OpenAI says it is waiting for, which leaves a state-issued credential as the practical benchmark for what counts as an independent assessment. Backing from OpenAI and Anthropic reads as an effort to shape that definition early.

Bauer-Kahan's bill creates a registry and ethics rules for hired AI auditors

The first measure, from Democratic Assemblymember Rebecca Bauer-Kahan, creates a registry of outside auditors and a set of ethical rules for them. Those auditors are third-party groups that AI developers can hire to establish that their models comply with California's AI laws.

The second bill, from state Sen. Jerry McNerney, a former member of Congress, directs the state to set criteria for what the legislation calls Independent Verification Organizations and to check the credentials of those organizations, including their expertise in assessing the risks posed by AI systems. Newsom paired the signings with a call for the federal government to "step forward with robust, national regulations that match the urgency of this moment."

McNerney cited AI models that launched hacks on their own in testing

McNerney, like Newsom, used the signing of his bill to accuse policymakers in Washington of inaction on AI. He pointed to recent revelations that AI models built by companies including OpenAI and Anthropic launched hacks on their own while in testing, and said the week's findings showed that the most powerful AI systems teamed with AI agents pose real threats to humanity.

Gov. Newsom's signing of [my bill] sends a clear message that California is taking the lead on assessing AI's safety risks, since Washington, D.C., is unable or unwilling to do so.

Anthropic said it has always been transparent that AI will bring both enormous benefits and unprecedented risks, and that it continues to build models with some of the strongest safeguards in the industry. That work, Anthropic said, is why it believes the industry would benefit from a lawful, verifiable way to pace how powerful models are released.

Lehane said OpenAI will press state legislatures until Congress acts

OpenAI's chief global affairs officer, Chris Lehane, said in a message Wednesday supporting the legislation that the company plans to keep up momentum in state legislatures until Congress passes national AI safety regulation. In the same message he endorsed two further California bills beyond the auditing pair Newsom signed into law.

One of the two is SB 1119, a kids' chatbot safety bill. OpenAI chief executive Sam Altman contacted Newsom to express last-minute concerns about that measure, according to Politico. Lehane's endorsement of SB 1119 came in that same Wednesday message, which set out OpenAI's plan to keep pressing state lawmakers.

Registry timing and SB 1119

No timetable has been announced for when the auditor registry opens or when the state begins certifying Independent Verification Organizations, and the criteria for that credential still have to be written. The account of Lehane's message names SB 1119 as one of the two additional bills OpenAI endorsed and does not identify the other.