Skip to content
IAM Insights

Explainers · Glossary

ERC-8004

ERC 8004

A proposed Ethereum token standard for an on-chain trust layer for AI agents: identity, reputation, and validation registries anchored to a blockchain so one organization's agents can be identified and vetted by another's.

In our briefings

The claim each of the newest briefings on ERC-8004 would have you leave with.

  1. Revision -15 closes a gap between delegation hops that the draft now describes itself: under -14, "anyone holding a copy of a parent receipt could mint a narrower child passing every chain check." A parent receipt can now name, in an optional delegateeKey field, the key allowed to sign beneath it. A sub-receipt's signing key must match that key, compared as RFC 7638 thumbprints, and a parent without the field is terminal, so issuing a sub-receipt under it must fail. The orchestrator signature that revision -13 had tied to the sub-receipt signer's key is deprecated. The change log calls this a breaking change: -14 chain test vectors without the field fail by design, -14-era verifiers reject receipts that carry it, and "There is no grandfathering of pre-15 chains." As we read it, the binding limits who can extend a chain. The draft still calls receipts bearer instruments that possession plus a valid signature authorizes, and only its optional OAuth sketch matches the presenting agent's key.

    Delegation receipts for AI agents, revision 15: holding a copy of a receipt no longer lets you extend the chainBI-FIND-2026-1116

  2. The risk both CISOs land on is manipulation rather than credential theft. An attacker who shapes what an agent reads gets it to misuse privileges it legitimately holds, the confused-deputy pattern, and that happens after identity and policy have done their jobs correctly. Gregorian's summary is that intent is missing from the security models: an agent can be authorized to act and have no judgment about whether it should, where a person would hesitate. The agentic-JWT author built a whole draft around the same hijack, and the summit panel called the intent-to-permission gap the one nobody has tooled.

    The agent risk two CISOs fear is manipulation, not a stolen credentialBI-FIND-2026-0127

  3. De Valois-Franklin and Bogdan argue that the agent-commerce protocols each answer an earlier question than the one disputes turn on. In their reading, AP2 establishes that the agent had authority, x402 that the payment can move, Visa's Trusted Agent Protocol that an authorized agent is present, and Mastercard's Verifiable Intent that a fulfillment event occurred, "not that the fulfillment satisfied the user's intent." Their worked case is a Verifiable Intent purchase where every credential checks out and a 65-watt charger arrives instead of the 100-watt one ordered. Their proposed fix, RAILS, is a clearing step that sits between those protocols and settlement and releases money only when the evidence meets a floor the deal sets in advance.

    Agent payment protocols prove an agent may pay; a Toronto paper adds a step to check the job was doneBI-FIND-2026-0899

›Cited in 9 briefings

Every briefing that discusses this term, newest first.

  1. Delegation receipts for AI agents, revision 15: holding a copy of a receipt no longer lets you extend the chain

    Draft · · IETF Datatracker · Ryan Nelson, Authproof · Oct 22, 2026

  2. The agent risk two CISOs fear is manipulation, not a stolen credential

    Podcast · · Defense in Depth · Identity and Access Management (IAM) in an Agentic AI World · Jul 23, 2026

  3. Agent payment protocols prove an agent may pay; a Toronto paper adds a step to check the job was done

    Paper · · arXiv · Adrian de Valois-Franklin et al. · Jun 7, 2026

  4. One x402 payment, replayed 1,000 times, was accepted up to 248 times; no audited SDK ties it to what it paid for

    Paper · · arXiv · Zelin Li et al. · May 12, 2026

  5. Most agent-payment protocols leave the agent's own reasoning out of scope, a survey of 142 works finds

    Paper · · arXiv · Qian'ang Mao et al. · Apr 15, 2026

  6. A January design paper argued AP2 and x402 move money before anyone checks the work, and proposed escrow

    Paper · · arXiv · Mehul Goenka et al. · Jan 30, 2026

  7. A Shenzhen paper ties each AI agent to a verified owner and its exact code, at 15 to 40 seconds of proving a step

    Paper · · arXiv · Zibin Lin et al. · Dec 19, 2025

  8. ERC-8004 stopped requiring agents to approve who rates them in January and admits fakes can inflate ratings

    Article · · Ethereum Improvement Proposals · Marco De Rossi et al. · Aug 13, 2025

  9. Decentralized AI agents have no off switch by default; builders say cutting API access still stops most of them

    Paper · · arXiv · Botao Amber Hu et al. · May 14, 2025

Related terms

What appears alongside ERC-8004 in the same briefings.

x402CoinbaseEIP-712