Definition
A security approach where access decisions for autonomous AI agents are made continuously and in real time based on current context, rather than relying on static credentials or one-time permission grants. Each agent action is evaluated against live risk signals, ownership, caller identity, and policy.
Why it matters
Traditional IT security assumes humans make occasional, supervised decisions. AI agents act continuously at machine speed. Continuous authorization replaces the old 'grant access and forget' model with real-time enforcement that adapts as risk changes.