Governor Gavin Newsom signs an EO directing a working group to provide a guide in two months with measures to boost California's AI safety and security laws

California could mandate 'kill switches' for AI models, providing a new compliance burden for AI developers.

California Governor Gavin Newsom has issued an executive order directing a working group to develop recommendations for the state's AI safety and security laws within two months. This initiative stems from growing concerns over the rapid advancement of artificial intelligence and perceived inaction at the federal level. Newsom stated that Washington has failed to provide meaningful oversight, prompting California to take a proactive stance. The proposed measures could include requirements for independent third-party safety plan reviews for frontier AI companies and the implementation of emergency shutdown mechanisms, often referred to as "kill switches," for AI models. This move positions California as a leader in AI regulation, building upon existing state laws that establish a framework for AI auditing. Other political figures, including Pennsylvania Governor Josh Shapiro and Senators Mark Kelly and Cory Booker, have also voiced concerns and proposed legislative actions regarding AI risks and worker displacement.

Governor Newsom's executive order signals a significant state-level push to regulate advanced AI, potentially setting a precedent for other states and even influencing future federal policy. By tasking a working group with creating a guide in just two months, California is prioritizing speed in its regulatory approach, emphasizing urgency over lengthy deliberation. This action directly addresses the growing anxieties about AI's potential risks, such as existential threats and societal disruption, and contrasts sharply with the federal government's slower pace. The governor's rhetoric suggests a desire to establish California's existing AI regulatory framework as the national baseline, potentially creating a patchwork of state-specific rules if federal action remains stalled.

The directive for potential mandates like third-party safety plan reviews and emergency shutdown capabilities indicates a focus on tangible, enforceable safety mechanisms for the most powerful AI systems. This approach could introduce substantial compliance costs and operational complexities for frontier AI companies operating within or marketing to California. It also highlights a strategic pivot towards proactive risk mitigation, moving beyond merely addressing potential harms to embedding safety requirements into the development lifecycle of cutting-edge AI. The speed requested for the working group's report suggests that these regulatory changes could be implemented relatively quickly, impacting the AI development landscape sooner than anticipated.