Definition
The mechanisms used to verify which AI agent or model is making a request, prove it has permission to do so, and hold it accountable for its actions — equivalent to the login credentials and access badges used for human employees, but designed for autonomous software agents acting at machine speed.
Why it matters
Existing identity systems were not built for agents that can spin up thousands of instances simultaneously, act for milliseconds, and never 'log in' in the traditional sense. Without proper agent identity, it is impossible to audit what an AI did, revoke its access, or detect when it has been compromised.