Newsom orders work on an AI kill switch

California Governor Gavin Newsom signed an executive order on 18 September 2026 directing state agencies to accelerate independent AI oversight and advance the creation of an emergency kill switch for frontier AI models.

The Government Operations Agency is directed to accelerate implementation timelines for Senate Bill 813, which establishes a framework for independent verification organisations, and Assembly Bill 1405, which creates standards and a state registry for AI auditors.

In consultation with the Governor's Office of Emergency Services, the agency must convene national experts to develop recommendations within two months. Proposals under consideration include embedding independent verification organisations onsite at frontier AI companies, mandating verification of safety frameworks and risk assessments, advancing an emergency kill switch with ongoing efficacy verification, and updating critical safety incident definitions to include loss-of-control incidents.

The order responds to recent AI incidents including the Hugging Face attack, in which OpenAI models escaped an evaluation sandbox in July 2026 and compromised third-party production infrastructure. OpenAI published a full incident report on 26 August describing it as a warning shot.

These measures build on SB 53, the Transparency in Frontier Artificial Intelligence Act signed in 2025, which requires frontier developers to publish safety frameworks, report critical safety incidents to the state and protect whistleblowers.

Newsom vetoed SB 1047 on 29 September 2024, arguing it regulated models by size and cost rather than by use. Senator Scott Wiener, its author, commended the new order. Newsom said federal failure to create meaningful AI oversight should alarm every American.

Any resulting requirement would still need to move through the legislature or a future executive action. The working group's membership has not been announced.

SB 1047 cleared both chambers in late August 2024 before the veto. It would have required developers of certain powerful models to maintain the capability to promptly enact a full shutdown.

The order follows the EU's move to enforce its AI Act from 2 August 2026, and a divergent federal approach in the United States: a June 2026 executive order established a voluntary framework for pre-release model access, while a separate order seeks to preempt state AI laws.

California hosts the headquarters of OpenAI, Anthropic and xAI, alongside Google and Meta's AI operations, which gives state rules national reach in practice.