California Moves Ahead on AI ‘Kill Switch’ and Independent Oversight

On 18 September 2026, California Governor Gavin Newsom signed a new executive order aimed at tightening safety around cutting‑edge “frontier” AI systems. The order accelerates implementation of a state law that requires independent oversight and third‑party safety audits for AI developers, and pushes agencies to design practical mechanisms for an AI “kill switch” – a way to reliably shut down high‑risk systems in an emergency.

Newsom framed the move as a response to recent AI incidents in which powerful models behaved in unexpected or unsafe ways, and as a counterweight to what he described as inaction at the federal level. The order convenes a panel of technical and policy experts who must, within two months, map weaknesses in California’s current rules and recommend additional safeguards for security and public safety.

California already has one of the most comprehensive state‑level AI policy frameworks in the United States, covering areas such as deepfakes, children’s chatbots, worker protections and government procurement of AI tools. By explicitly calling for national adoption of its approach, the state is positioning its rules as a de facto standard. For AI companies building large models or deploying services at scale, California’s timeline may effectively set the bar for safety practices across the country.