ARIA: the trust handshake that lets any organisation verify an AI agent at its door

How it works

1. An operator registers its agent with the ARIA Registry and names its Principal — the legal entity accountable for what that agent does.

2. The registry issues an AID: a W3C Verifiable Credential naming the agent, its Principal and its trust level, signed with a post-quantum composite suite. It is held by the operator's systems — the holder here is a machine, not a person.

3. The receiving organisation publishes its own ARIA Agent Trust Policy as a DNS TXT record on its domain: the minimum trust level it accepts, and its enforcement mode — monitor, warn or strict.

4. The agent arrives and DECLARES its credential and its intent.

5. The receiver EVALUATES both against the policy it published itself, verifying the signatures locally with the open-source verifier.

6. The receiver ADMITS or refuses, returning a defined response code — so a refusal is auditable and machine-readable, never a silent drop.

7. If the credential is later suspended or revoked, the event is written to an append-only ledger and reaches verifiers in about sixty seconds. The same agent knocking again is refused, at every door, without anyone being notified by hand.

Tags

• agent identity
• AI agents
• organisational identity
• verifiable credentials
• decentralized identifiers
• post-quantum cryptography
• DNS
• revocation
• accountability
• open standard
• nonprofit governance
• agentic commerce

Interaction modes

• Remote flow

VC formats

• VCDM2.0

Issuers involved

• ARIA Registry

Credential types used

• AID — Agent Identity Document

Relying parties

• https://www.aria.bar/spec