AI Agent Risk Assessment

A booking bot and an agent that can alter financial records should not get the same review. This tool scores one agent from its authorized actions, autonomy, oversight, logging and the data it touches.

Scoring

Action impact (highest-impact authorized action) × autonomy multiplier × oversight adjustment, with additions for sensitive data, external communication and missing logs. Tiers: Low, Elevated, High, Critical. It is an internal prioritization model, shown in full, not a legal classification. Read AI agent governance explained and Human oversight for autonomous agents.