Interruptibility is the ability of an autonomous system to accept human override, pause, or shut down safely when conditions change. In robotics and agentic AI, it means the machine can yield control instead of blindly continuing toward a goal. This is a core safety property because a system that cannot be stopped on request, or that resists interruption, can turn a small mistake into physical harm or operational damage.
In cyber security, interruptibility matters because attackers may try to disrupt stop logic, spoof sensors, flood a controller, or exploit software flaws that prevent a safe halt. Defenders design emergency-stop paths, monitored shutdowns, force and speed limits, and runtime supervision so humans can intervene when uncertainty rises. Good interruptibility also supports incident response: if a robot, drone, or AI agent behaves unexpectedly, operators need a reliable way to pause it, isolate it, and restore control without making the situation worse.