US lawmakers are considering a requirement for AI systems to include “kill switches” that could shut them down if they become too dangerous. The proposal would focus on emergency shutdown mechanisms for advanced AI systems, though no further details were provided.