Anthropicโs updated Usage Policy takes effect on 12 November 2026, introducing stricter safety requirements for equipment that uses Claude to perform autonomous physical actions capable of causing injury. Under the new rules, a qualified human operator must be able to monitor the equipment and intervene at any time. If the operator intervenes or the connection to Anthropicโs services is lost, the equipment must stop or move into a defined safe state. Crucially, safety limits governing parameters such as speed, force, temperature and pressure must be enforced by the machine itself or an independent controllerโnot solely by the AI model. The requirements cover relevant applications involving industrial machinery, mobile robots, hazardous materials, safety systems and other potentially dangerous operations. This development highlights an important principle for the future of industrial AI and robotics: autonomous intelligence must operate within independently enforced safety boundaries, with meaningful human oversight and fail-safe mechanisms.
