Runtime Authorization

A security control mechanism that requires an AI agent to request and obtain verified permissions in real time before each high-sensitivity operation, rather than receiving broad permanent permissions at deployment, thereby dynamically limiting the potential damage radius.