Newsom orders California to study an AI kill switch by November

California governor Gavin Newsom signed an executive order on 18 September 2026 giving state officials until 16 November 2026 to recommend changes to the state’s AI safety laws, including whether to require a kill switch for the most capable models.

The order commissions a study rather than a rule, which is the limit worth stating plainly. Its significance is that it moves ideas that circulated as industry talking points, independent verification and mandated shutdown capability, into a state process with a deadline.

What the order actually requires

The order directs officials, working with national experts, to return recommendations by 16 November 2026. Four proposals are named for consideration. The announcement asks officials to “advance the creation of a ‘kill switch’ for frontier models, with the efficacy of the switch verified on an ongoing basis by an independent verification organization,” to “require frontier AI companies to embed a designated independent verification organization onsite in their labs to conduct regular audits and evaluations,” to have filings required under state law verified against independent standards, and to “update definitions of critical safety incidents to include loss-of-control incidents.”

None of those takes effect on signature. The order asks whether each should be required.

Newsom framed it against federal inaction. “The federal government’s abject failure to create any form of meaningful AI oversight or accountability should alarm every American, especially when AI CEOs themselves are begging for regulation,” he said. He also set the pace as the point: “We’re not waiting to act, we’re going to speed up our work on substantial and responsible AI oversight before it’s too late.”

The ideas were already in circulation

The kill switch question had been raised from inside the industry. Anthropic co-founder Jack Clark asked on the BBC whether companies should be required to have one, and whether its operation should be verifiable by a third party. The order puts both of those to study.

In Congress, the AI Kill Switch Act from Representatives Ted Lieu and Nathaniel Moran was introduced in July 2026 and has not advanced. That leaves California’s November deadline as the nearer of the two dates on which anything concrete could emerge. The signed order names 16 November 2026; the announcement describes it as within two months.

What remains open

The hard part is verification, not the mandate. A shutdown capability a lab attests to is a policy; one an outside body confirms on an ongoing basis is an engineering and access problem nobody has solved at frontier scale, and the order does not say how such an organization would be constituted or what it could see.

The recommendations due on 16 November 2026 are where that question gets an answer or does not.