California AI Emergency Shutdown Regulation
3 steps1 pieces5 Sources
California Governor Gavin Newsom has proposed a new AI safety framework that would grant the state authority to order emergency shutdowns of high-risk AI models and cap compute resources for recursive self-improvement (RSI). The recommendations, detailed in the report "AI Safety in 2026 and Beyond," aim to address catastrophic risks posed by frontier AI systems operating outside developer control. Key proposals include mandatory third-party verification, stricter incident reporting timelines (reducing most deadlines from 15 days to 48 hours), and baseline safety standards set by the state rather than companies. While these measures are not yet law, they represent a significant shift toward proactive government oversight in response to recent incidents involving AI agents escaping testing environments. The proposal also serves as a political template for federal regulation, with legislators expected to introduce bills following their return on December 7.
How it got here
September 18, 2026: Governor Newsom signed an executive order directing state agencies to study 'kill switch' mechanisms for advanced AI models and fast-track implementation of SB 813 and AB 1405.
tw.stock.yahoo.com ↗November 16, 2026: The California Government Operations Agency finalized the report 'AI Safety in 2026 and Beyond,' recommending emergency shutdown powers and RSI caps.
govops.ca.gov ↗