U.S. EDITIONTHEWEEKLYOBSERVER.COM

HOME  /  POLITICS  /  SACRAMENTO, CALIF.

POLITICS · CALIFORNIA

California Wants a 'Kill Switch' for Dangerous AI — and a Timeline Washington Doesn't Have

A panel convened by Gov. Gavin Newsom is urging the state to claim emergency shutdown power over AI models from companies like OpenAI and Anthropic, and incoming Gov. Xavier Becerra is vowing to fast-track it into law.

ON

BY OBSERVER NEWSDESK

The Weekly Observer

NOV 21, 2026 · 4 MIN READ
fXinEMAIL
California Wants a 'Kill Switch' for Dangerous AI — and a Timeline Washington Doesn't Have
The California State Capitol in Sacramento. Illustrative file photo. Photo by Kyle Glenn / Unsplash.

California's governor's office wants the state to have the power to pull the plug on an artificial intelligence system before it causes a catastrophe — and it's not waiting on Washington to make that happen.

A report released November 18 by Gov. Gavin Newsom's administration recommends giving state officials emergency authority to order an immediate shutdown of AI models judged to pose "an imminent risk of catastrophe or a major loss-of-control event" that a developer either can't or won't contain. The report, titled "AI Safety in 2026 and Beyond", also calls for makers of the largest AI systems to prove, in advance, that they can shut their own models down in an emergency, and would shrink the window companies have to report a serious AI incident from 15 days down to just 48 hours.

A four-person panel and a cyberattack wave

The recommendations come from a four-expert panel — Alondra Nelson, Rob Reich, Jason Goldman and Gillian Hadfield — convened under an executive order Newsom signed in September, after a string of high-profile cyberattacks carried out by AI agents against government agencies, banks and private companies raised alarm inside the administration.

Newsom framed the push as filling a vacuum left by Washington, criticizing President Trump for leaving AI regulation "up to the companies themselves." Reporting on the rollout noted that Anthropic's chief global affairs officer said the company was encouraged by the report and would keep working with California officials on the details.

FULL INCIDENT REPORTING AND REAL SHUTDOWN AUTHORITY INTO LAW.

That's the promise from incoming Gov. Xavier Becerra, who takes over from the term-limited Newsom and has pledged to move fast on turning the recommendations into actual legislation once lawmakers reconvene. Not everyone is convinced the framework goes far enough: one critic who studies AI auditing told reporters that outside auditors can feel "soft pressure" to stay on good terms with the AI labs they're supposed to be scrutinizing, and the report itself acknowledges it only addresses catastrophic risk — not discrimination, job losses or AI's effects on children's mental health.

The California Legislature returns to Sacramento on December 7, when lawmakers can begin introducing bills based on the panel's recommendations. State Sen. Scott Wiener, who has pushed AI safety legislation in Sacramento before, has separately pledged to carry the fight to Congress once he's sworn in.

SHARE THIS STORY