Newsom executive order puts AI kill switch mandate and onsite audits on California's agenda

California Governor Gavin Newsom has issued an executive order positioning the state to take the lead on AI oversight, including a possible mandate for a "kill switch" on frontier models. According to The Verge, the order was signed on Friday and directs California to convene a group of experts who must deliver recommendations within two months on how to strengthen AI safety measures in state law.

What the order puts on the table

The Verge reports that Newsom wants the expert group to evaluate a specific set of mechanisms:

• Requiring AI companies to embed independent verification groups onsite for regular audits
• Making company transparency reports and risk assessments subject to the standards of independent auditors
• Creating a kill switch that is routinely verified as effective
• Obligating companies to report "loss-of-control incidents" as critical safety events, with The Verge citing an OpenAI attack on Hugging Face as the kind of episode Newsom has in mind

The order also directs a state agency to speed up implementation of two laws Newsom recently signed: one establishing a framework for independent verifiers to assess AI safety, and another creating a state registry of AI auditors.

California moves while Washington stalls

Congress has seen a burst of AI bill discussions in the wake of a post by former Anthropic researcher Jacob Coxon warning of a more than 10 percent chance that AI "could kill all humans" within a decade, The Verge reports. But the outlet sees little prospect of substantive federal legislation any time soon: the House is in recess until November, and the approaching midterm elections make major movement in the Senate similarly unlikely.

President Donald Trump has dismissed the need for formal rules, insisting that the only "guardrails" AI requires is a strong president. Newsom is pitching California's work as an alternative, calling on Congress and Trump to "review and adopt the state's framework," or to use its regulation as a floor rather than a ceiling, according to The Verge.

In a statement, Newsom said the federal government's failure to create any form of meaningful AI oversight or accountability should alarm every American, particularly given that AI executives themselves are asking for regulation, and said the state would proceed with urgent velocity rather than wait.

A special session may be next

The executive order is probably not California's last word on the subject. The Verge notes that Newsom told Politico on Wednesday that he is considering calling state lawmakers into a special legislative session focused on AI, alongside potential further executive action.

Why it matters

If the expert group's recommendations become law, developers of frontier models, which in practice means most large AI labs, many of them headquartered in California, could face binding requirements for onsite audits, independently reviewed risk assessments, mandatory loss-of-control incident reporting and a shutdown mechanism proven to work. Practices that exist mostly as voluntary commitments today would become enforceable obligations, with real compliance costs and engineering implications.

The move also illustrates how states are stepping into the vacuum left by federal inaction, and Newsom is explicitly inviting the federal government to treat California's framework as a national baseline. With recommendations due in two months and a special legislative session on the table, concrete bill language could arrive quickly. Companies building or deploying frontier models may want to begin planning for auditor access, incident reporting pipelines and tested shutdown controls well before anything is signed into law.