California Pushes AI ‘Kill Switch’ Oversight
California Governor Gavin Newsom has signed an executive order accelerating independent oversight of artificial intelligence companies and directing experts to examine stronger safeguards for advanced AI systems, including a potential emergency “kill switch” for frontier models.
California Explores AI Kill Switch for Frontier Models
The executive order does not immediately require AI companies to install a kill switch. Instead, it directs California officials to convene experts who will recommend within two months how the state can strengthen its existing AI safety and security framework.
Among the proposals to be examined is a requirement for frontier AI developers to create an emergency mechanism capable of shutting down advanced models when serious safety or loss-of-control risks emerge. The effectiveness of such systems could also be independently verified on an ongoing basis.
Independent Audits of AI Companies Proposed
California is also considering requirements for independent verification organisations to operate within frontier AI laboratories and regularly assess their systems.
Safety frameworks, transparency reports and risk assessments submitted by AI developers could be subjected to third-party verification under the proposed approach.
Newsom’s order additionally calls for the definition of critical AI safety incidents to be expanded to include situations involving potential loss of control over advanced models.
California Accelerates Existing AI Safety Laws
The order directs state agencies to speed up implementation of two recently enacted laws, SB 813 and AB 1405.
SB 813 establishes a framework for independent organisations capable of evaluating AI systems for safety and risk, while AB 1405 creates a state registry for AI auditors and sets standards covering independence, transparency and integrity.
Newsom has also urged the federal government to consider California’s regulatory framework as a baseline for wider national AI safety standards.
