Gavin Newsom is pushing for an AI kill switch

California Governor Gavin Newsom has signed an executive order directing state officials to convene a panel of experts to produce recommendations within two months on strengthening AI safety in state law. The order asks the group to evaluate a range of measures aimed at increasing oversight of so-called frontier AI models and to recommend concrete legal and regulatory steps for the state to take. Among the potential requirements Newsom wants considered are mandating that AI firms host independent verification teams onsite for routine audits, subjecting company transparency reports and risk assessments to independent-auditor standards, and developing a "kill switch" for frontier models that would be routinely verified as effective. The order would also make reporting of "loss-of-control incidents"—the executive cites the OpenAI incident affecting Hugging Face as an example—into the category of critical safety incidents. The executive order further instructs a state agency to accelerate implementation of two recent laws Newsom signed that establish a framework for independent AI verifiers and create a state registry of AI auditors. Newsom has portrayed the state effort as a potential model for federal action, calling on Congress and President Trump to review and adopt California’s framework or use it as a regulatory floor. Newsom has also suggested the possibility of calling a special legislative session in California on AI. His action comes amid heightened public and legislative debate over AI risks following warnings from researchers — including a former Anthropic researcher’s post estimating nonzero existential risk — and at a moment when major federal legislation appears unlikely to advance quickly because Congress is in recess and midterm politics loom.