Summary
California Governor Gavin Newsom has issued an executive order to accelerate independent oversight of AI companies and develop recommendations for emergency shutoff mechanisms for frontier AI models. A national expert group is expected to produce a guide for strengthening the state's AI safety laws within two months.
California Governor Gavin Newsom has issued an executive order directing the state to accelerate independent oversight of AI companies and advance proposals for an emergency shutoff mechanism—described in the announcement as an “AI kill switch”—for frontier AI models.
The order asks a group of national experts to prepare, within two months, a guide for strengthening California’s AI safety and security laws. The proposed changes include independent audits, verification of company safety disclosures and ongoing checks that an emergency shutoff mechanism works as intended.
What the executive order directs
The order directs California’s Government Operations Agency, in consultation with the Governor’s Office of Emergency Services, to speed up implementation of two laws signed in 2026: Senate Bill 813 and Assembly Bill 1405.
SB 813 established a framework for certifying independent verification organisations with the expertise and independence needed to assess AI systems and models for safety and risk. AB 1405 created a state registry for AI auditors and set standards covering their independence, transparency and integrity.
The new order asks experts to recommend whether California should require frontier AI companies to:
- have a designated independent verification organisation conduct regular audits and evaluations onsite in their laboratories;
- have safety frameworks, transparency reports and risk assessments verified against standards judged adequate by an independent verification organisation; and
- develop an emergency shutoff mechanism for frontier models, with its effectiveness checked on an ongoing basis by an independent verification organisation.
The experts are also being asked to recommend expanding the definition of critical safety incidents to include loss-of-control incidents. The announcement cites a recent incident involving Hugging Face as part of the context for the order.
The “kill switch” proposal therefore refers to a policy requirement for frontier-model developers to create and maintain a mechanism capable of shutting down a model in an emergency. The announcement does not specify the mechanism’s engineering design, who would have authority to activate it or the conditions that would trigger it.
A proposed layer of independent scrutiny
California’s approach would place outside assessment between AI companies and state oversight. Instead of relying only on a company’s own safety documentation, independent verification organisations could be involved in evaluating models, reviewing safety plans and checking whether emergency controls function over time.
The state’s existing framework, as described in the executive-order announcement, already includes a 2025 law requiring frontier AI developers to publicly disclose safety frameworks, report specified critical safety incidents and protect whistleblowers who report serious risks. The 2026 laws add institutional arrangements for independent verification organisations and registered AI auditors.
The executive order accelerates those implementation timelines and asks experts to propose further changes. The emergency shutoff is consequently at the recommendation and implementation stage rather than being described as an already operating statewide system. Any final requirements would depend on the expert recommendations, subsequent state action and the technical standards adopted for verification.
The order also presents California’s framework as a possible model for national AI standards. Its immediate legal effect is within California, while the call for national adoption is a policy request to Congress and the federal government.