A stop rule is a predefined condition for pausing or suspending a system when uncertainty, risk, or malfunction exceeds an acceptable threshold. It turns a general safety concern into a clear operational trigger, helping users avoid continuing to rely on a tool when its output may be unsafe.
In cyber security, stop rules can apply to automated monitoring, incident response, or AI-assisted decisions. For example, a workflow might pause automated actions when a system detects conflicting signals, loses required safeguards, or behaves unexpectedly, then escalate to human review or a documented fallback. In clinical AI, a stop rule can help preserve independent judgment by making it easier to step away from automation when uncertainty or risk rises. The rule should specify who can pause the system and how work continues safely.