Anthropic co‑founder calls for mandatory AI kill switch

Jack Clark, a co‑founder of the AI safety startup Anthropic and former policy director at OpenAI, warned that while most artificial‑intelligence research labs already possess various mechanisms to halt their own systems, the industry lacks a uniform standard for such “pull‑the‑plug” capabilities. In a recent interview, Clark noted that the diversity of shutdown procedures across labs reflects differing technical architectures and risk‑management cultures, but he argued that a consistent, enforceable requirement could improve overall safety oversight.

Clark suggested that regulators and industry bodies consider mandating a baseline shutdown protocol for advanced AI projects, ensuring that any system deemed unsafe can be reliably disabled regardless of its internal design. He emphasized that a standardized requirement would help align practices across the sector and provide clearer accountability, potentially reducing the risk of uncontrolled or harmful AI behavior. The proposal is expected to spark further discussion among policymakers, researchers, and corporate leaders about the balance between innovation and safety safeguards.