Who is it?
Identity & Provenance
Binds an agent to a real, accountable party. Answers who — not what the agent did.
The AI agent accountability landscape
As AI agents start acting on their own, a whole category is forming around one question: can you trust what an agent did? Here is how the space breaks down — and where each piece fits.
who is it? → what's it allowed to do? → what did it actually do? → what's its track record? → by what rules?
What's it allowed to do?
Governance & Guardrails
Policies and gates that decide, before execution, what an agent may do. Prevention, not proof.
What did it actually do?
Evidence & Receipts
Signed, tamper-evident, independently verifiable records of each action. Proof, not just policy — the half most of the stack skips.
What's its track record?
Trust Scoring & Reputation
Turns a verified history of behavior into a portable score — a "credit score" for agents.
By what rules?
Standards & Regulation
The specs and laws defining what a record must prove and when it's required.
How to read it: identity and governance answer who and what's allowed; evidence answers what actually happened and lets a third party verify it without trusting the operator; trust scoring turns that history into reputation; standards set the rules. They're complementary — but the evidence layer is the one most teams skip. Full sourced list: awesome-ai-agent-accountability. Inclusion is by relevance, not endorsement.