News

Gavin Newsom Demands California Design an AI Kill Switch After Vetoing That Exact Law

Gavin Newsom signed an executive order creating a panel to recommend AI safety rules, including a possible kill switch, within about two months. He vetoed a shutdown bill in 2024.

Gavin Newsom Demands California Design an AI Kill Switch After Vetoing That Exact Law

Gavin Newsom wants California to design an AI kill switch. Two years ago he vetoed a bill that would have required one.

The governor signed an executive order on Friday, 18 September 2026, standing up a panel to recommend tougher AI safety rules, including a possible shutoff for programs that go rogue, the Los Angeles Times reported. The working group has about two months to report. The order also asks whether independent third parties should write the safety plans.

Labs do not have to install a working kill switch today. Washington has not passed a matching federal law.

Newsom killed SB 1047, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, on 29 September 2024. That bill would have forced developers of covered models to "promptly enact a full shutdown," stopping both operation and further training. He said then that it regulated by model size and cost instead of by use.

Now he says a kill switch "means a lot of things depending on who you talk to," and that the panel should pin the phrase down. The order landed a day after Pennsylvania Governor Josh Shapiro made a similar safeguard plea, in a week already thick with catastrophe warnings from employees and experts.

On his podcast, Newsom noted there is still no comprehensive federal law requiring AI companies to report dangerous incidents, and that President Trump has rejected new rules. Newsom boiled that stance down to "Let it rip" and said Trump is "dead wrong." Trump has called apocalyptic AI warnings a hoax. That does not settle the engineering argument.

TechNet said it will work with Newsom. Nvidia CEO Jensen Huang has said no new regulation is needed and that safety is an engineering hold-back problem. Anthropic CEO Dario Amodei has argued for pacing the frontier, the same split sitting under the Anthropic and OpenAI pace fight. Beijing, for its part, has called slowdown warnings fear mongering.

Senator Christopher Cabaldon is planning fall hearings. He said California already regulates a barber or a sandwich shop more tightly than a company that says it may be ending humanity. Senator Scott Wiener, who wrote the 2024 bill, called the order a challenge to pass strong guardrails early next year.

Europe already has a different kind of off switch. Since 2 August, the European Commission can restrict, withdraw, or recall a general-purpose model under the AI Act. That power sits with the regulator, not a button inside the lab. Geoffrey Hinton told CNN a company kill switch may fail because AI would "persuade the people in charge of the switch not to pull the switch."

Friday's order starts a two-month clock. It does not flip one.

Some offers on this page may be paid placements or contain affiliate links.

Subscribe to Techpresso

Free daily newsletter, read in 5 minutes.

Subscribe free