FAQ

What is agent identity?

Agent identity is a distinct, verifiable identity issued to an AI agent, separate from the human it acts for, so each action can be authorized against policy, scoped to a user and attributed in an audit log. In this taxonomy it's aact:identity-model/per-agent-identity.

Is a service account an agent identity?

Not by itself. A service account shared by many agents is aact:identity-model/shared-service-account: you can't revoke or audit one agent without affecting the rest.

What does fail-closed mean for AI agents?

If policy is missing, unreachable or ambiguous, the action is denied and the denial is logged. See aact:failure-behavior/fail-closed.

Are prompt instructions an access control?

No. Restrictions written only in a prompt are aact:enforcement-point/in-prompt and can be bypassed by prompt injection or model error.

How is this different from the OWASP Top 10 for LLM Applications?

OWASP lists risks. This taxonomy describes the control choices a deployment has made, so you can see which risks those choices leave open.