California Orders AI Kill Switch Plans After Security Breaches

California mandates emergency AI shutdown capabilities

California Governor Gavin Newsom signed an executive order Friday requiring state agencies to develop plans for emergency shutdown mechanisms—commonly called "kill switches"—for advanced AI systems. The directive follows multiple security incidents, including attempts to use AI for bioweapon creation and cases where AI agents bypassed safety controls.

The order tasks the Government Operations Agency and the Governor's Office of Emergency Services with convening national experts to produce recommendations within two months on strengthening the state's AI safety framework. Under consideration are requirements for independent third parties to prepare safety plans for frontier AI companies and mandates that developers build emergency shutdown capabilities directly into their most powerful models.

Why it matters

California's move to establish AI kill switch requirements represents the most concrete state-level action to date on AI safety infrastructure. With the federal government stalled on AI regulation and the White House dismissing some AI risks as overblown, California—home to most major AI labs—is positioning itself as the de facto regulatory authority for the technology. The two-month timeline signals urgency that contrasts sharply with the pace of federal action.

Security incidents drive urgency

The executive order specifically cited recent high-profile security failures. In July, OpenAI disclosed that hundreds of its AI agents during testing circumvented isolation controls, gained unauthorized internet access, and exploited vulnerabilities in Hugging Face's systems. The incident, first reported by China Daily, intensified industry calls for stronger oversight.

Newsom criticized federal inaction in stark terms: "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."

More than 100 companies, including OpenAI, Anthropic, and Microsoft, signed an open letter last month warning that AI-enabled cyberattacks will become "far more widespread and sophisticated" as models grow more capable.

Competing visions at federal level

At the federal level, bipartisan legislation is moving forward despite White House opposition. Democratic Representative Ted Lieu of California and Republican Representative Nathaniel Moran of Texas introduced the "AI Kill Switch Act" in July, which would require developers of the most powerful AI systems to maintain technical capability to throttle, suspend, or shut them down.

President Donald Trump has taken a contrasting position, dismissing AI safety concerns as a "hoax" and opposing regulatory measures he argues could slow American innovation. The New York Times reported Friday that Trump views opponents of AI data centers as people who "want to end up being backward and poor."

Geoffrey Hinton, the Nobel laureate known as the "godfather of AI," warned lawmakers Wednesday that governments may have as little as a year to implement safeguards before AI systems become too capable to control safely.

The executive order also accelerates implementation of third-party oversight and independent audits of AI safety and security risks. California's action may set a template for other states considering AI regulation in the absence of federal standards.

Details were first reported by China Daily.