C2PA
Coalition for Content Provenance and Authenticity
A content-provenance standard that attaches tamper-evident metadata to digital files so you can check where they came from and how they were changed.
In our briefings
The claim each of the newest briefings on C2PA would have you leave with.
Figuera's Sello protocol moves the record of an AI agent's actions from the agent to the services it calls. Each Sello-aware service seals a receipt body with HPKE to an owner key that the agent's JWS authorization token carries as a signed claim (owner_hpke_pk), signs the envelope with its own Ed25519 key and submits it to a public transparency log, where the owner finds it by the token's SHA-256 hash. In the paper's words, "The agent never holds any of the signing keys involved." The design assumes at least one service the agent calls is honest and uncompromised within the relevant time window: if every service colludes with the agent's operator, Figuera says, Sello cannot detect what they fabricate.
Sello has the services an AI agent calls sign its receipts; the agent can't fake them, but it can go around themBI-FIND-2026-2275
Kamimura's draft covers requests to generate content, not agent actions, and rests its audit on one rule. The AI system creates an ATTEMPT record before its safety check begins, and every ATTEMPT registered with a SCITT transparency service must have exactly one Outcome (DENY, GENERATE or ERROR) pointing back to it, so an ATTEMPT with no Outcome shows an auditor incomplete logging or a request still in progress. A request never logged leaves nothing to pair, the draft concedes. Removing an ATTEMPT together with its Outcome still satisfies the rule, and catching that takes a commitment to the whole set, such as the log itself. The transparency service should not try to enforce the rule. The threat model trusts auditors to check it, and leaves out of scope an AI system that bypasses logging or colludes with the transparency service. Even a logged refusal shows only that a refusal was claimed and when, not that it was correct. In the author's words, the scheme offers "auditability of refusal decisions that are logged, not cryptographic proof that no unlogged generation occurred."
A draft makes an AI system's refusals auditable in a SCITT log, but only for requests the system loggedBI-FIND-2026-1829
Few agents in the Index give a website anything stable to check. Only 7 of the 30 publish stable User-Agent strings and IP address ranges, while 6 explicitly use Chrome-like User-Agent strings and residential or local IP addresses, which the authors describe as mimicking human web traffic. ChatGPT Agent is the one agent they found signing its requests with HTTP Message Signatures (RFC 9421), so a receiving site can check each request against a published public key. Without such signing, the authors write, it is "significantly harder to verify agent identity or to prove what an agent actually did." The counts reflect public documentation as of 31 December 2025.
Of 30 deployed AI agents, 7 publish how a website can recognize them and one signs its requestsBI-FIND-2026-1430
›Cited in 9 briefings
Every briefing that discusses this term, newest first.
- Sello has the services an AI agent calls sign its receipts; the agent can't fake them, but it can go around them
Paper · · arXiv · Juan Figuera · Jun 2, 2026
- A draft makes an AI system's refusals auditable in a SCITT log, but only for requests the system logged
Draft · · IETF Datatracker · Tokachi Kamimura, VeritasChain Standards Organization · Oct 24, 2026
- Of 30 deployed AI agents, 7 publish how a website can recognize them and one signs its requests
Paper · · arXiv · Leon Staufer et al. · Feb 19, 2026
- A survey rates only workload identity ready for agents, and argues no standard can tell what an agent intends
Paper · · arXiv · Takumi Otsuka et al. · Apr 25, 2026
- OAuth handles an agent in one company, working while you watch. The OpenID paper says that is where it stops.
Article · · OpenID Foundation · Tobin South et al. · Oct 2025
- AI governance researchers argue OAuth tokens can't link an agent's work across services or sub-agents
Paper · · arXiv · Alan Chan et al. · Jun 19, 2025
- The 2024 'IDs for AI' paper puts the ID on a chat session, not a model, and leaves trust in its issuer to others
Paper · · arXiv · Alan Chan et al. · Jun 17, 2024
- In March 2024 ENISA found deepfake injection leading attacks on ID proofing, with no agreed standard to detect it
Article · · ENISA · Evgenia Nikolouzou et al. · Mar 12, 2024
- A 32-author 2024 paper asks for a credential that proves you are human and nothing more
Paper · · arXiv · Steven Adler et al. · Aug 15, 2024