Deals & regulation
California’s Newsom orders new review of AI kill switches and model safety rules
California Governor Gavin Newsom has ordered an independent review of AI models and created an expert committee to examine shutdown mechanisms for potentially dangerous AI agents. The move revisits measures he vetoed in 2024 while expanding the state’s existing frontier-AI reporting framework. Supporters say accelerating incidents show that industry self-regulation is inadequate; critics previously warned stringent rules could drive companies away. Analysts also view the order through the lens of Newsom’s expected 2028 ambitions. Implementation remains uncertain, with recommendations due November 15 and possible legislative action before his term ends.
Editorial summary · Publisher article
What matters
- The executive order asks for independent verification of AI models and recommendations on kill switches.
- It follows California’s 2025 frontier-AI law requiring large developers to disclose specified severe safety incidents.
- Experts say a shutdown mechanism is narrower and weaker than the vetoed bill’s proposed pre-training obligations.
- California may convene a special legislative session, but the governor’s remaining time in office is limited.
A narrower revival of an earlier proposal
The new order focuses on shutdown or stoppage mechanisms, while the earlier bill proposed controls before training runs and across a broader safety architecture. Critics therefore describe the order as less comprehensive than the vetoed legislation.
Policy and political uncertainty
The state’s emergency-services office is expected to provide recommendations by November 15, but it is unclear whether legislation will follow. Analysts quoted in the report also interpret the announcement as positioning against federal self-regulation and possible Democratic rivals.
Context & caveats
- The report includes competing assessments of Newsom’s motives and relies partly on claims about recent AI incidents that are not independently detailed in the packet.
Why this made the edition
The report gives concrete details of California’s new AI-safety actions, their relationship to the governor’s earlier veto, competing expert and industry views, and the political and implementation uncertainty.
Source material