Skip to content
IAM Insights

Explainers · Glossary

AI Agent Authentication and Authorization

draft-klrc-aiagent-auth · draft-klrc-aiagent-auth-01

A proposed IETF Internet-Draft on authentication and authorization patterns for AI agents acting on a user's behalf, since replaced by the WIMSE working group's draft-ietf-wimse-aims. A draft, not a standard.

reference →

In our briefings

The claim each of the newest briefings on AI Agent Authentication and Authorization would have you leave with.

  1. Uber runs per-hop token exchange with a full actor chain for AI agents in production, and has published sample tokens and a latency chart. Its security token service gives each hop a fresh token scoped to one audience, with a lifetime "in the order of minutes" (both sample payloads expire 300 seconds after issue), and the token carries the attested chain from the engineer to the agent making the call. Uber says "the system has been adopted by thousands of internal agents" and reports the exchange API's P99 as "consistently below 40 milliseconds" under current load; its chart sits mostly under 10 ms, with one spike reaching 40. The post covers this identity layer. Uber frames dynamic access control and a unified policy enforcement plane as its direction on top of it, and names session and intent claims as additions the token could carry in the future.

    Uber runs per-hop token exchange with a full actor chain for AI agents in productionBI-FIND-2026-0081

  2. Revision -04 of Hassan's draft reverses -03 on calls that change who can act on what. A call that grants or widens who can do what to which resource is now x, the execute class, "because someone can use the new access before it is taken back", for example an access policy, a role or permission binding, an access control list, a share or an invitation, or a deploy key. A call that only narrows access is w (write), and one that sets or replaces access without saying which way it goes is x. Section 17 is corrected to match: setting another person's password, or granting or widening their role or access, is x, where -03 put any change to another person's role, password or access in w, which a link that allows w admits. Where no signed menu names the call, though, the HTTP method sets the class, and PUT and PATCH default to w; the draft concedes that some of those calls are really x.

    Hassan's agent delegation draft now classes granting access as execute, not writeBI-FIND-2026-2333

  3. Revision -02 adds a section on what a server holding the credentials cannot promise. Compromise of the Delegation Server (DS), its operator or the path that releases the keys protecting stored credentials lets an attacker use those credentials with no Delegation Token at all, and Sweeney writes that "Keeping credentials away from Agents does not remove this exposure; it concentrates it at the DS." The server's signed receipts prove it authorized an action under its own records, but, in his words, "it does not prove that the Subject, rather than the operator, approved the grant". Approval signed with a key the user holds is left to a later revision. The revocation section now states its limits too. The one-second and five-second targets stand, yet "revoked_count reports invalidation, not closure": a token the server exchanged for direct use at a resource server, or issued under the offline profile, can be honored by an enforcement point off the server's path until it expires, at most an hour after issue, unless that point checks authority or status. Synchronous cascading revocation holds within one Delegation Server; enforcing it across servers or domains is out of scope.

    Sweeney's delegation draft concedes its server is trusted with users' credentials, and drops an interop claimBI-FIND-2026-1939

›Cited in 9 briefings

Every briefing that discusses this term, newest first.

  1. Uber runs per-hop token exchange with a full actor chain for AI agents in production

    Article · · Uber Engineering blog · Matt Mathew et al. · May 21, 2026

  2. Hassan's agent delegation draft now classes granting access as execute, not write

    Draft · · IETF Datatracker · Amr Hassan · Oct 22, 2026

  3. Sweeney's delegation draft concedes its server is trusted with users' credentials, and drops an interop claim

    Draft · · IETF Datatracker · Kieran Sweeney · Oct 13, 2026

  4. A post-quantum agent certificate draft warns that its own declared scopes are not proof of behaviour

    Draft · · IETF Datatracker · Reuben Burls · Jul 21, 2026

  5. Schrock's receipts prove a named key approved one exact agent action, and the draft lists at length what they cannot prove

    Draft · · IETF Datatracker · Iman Schrock, EMILIA Protocol · Sep 12, 2026

  6. Munoz's agent Permit is signed before the action runs, but the profile does not make the action wait for it

    Draft · · IETF Datatracker · Christian Munoz, Keel API, Inc. · Jul 19, 2026

  7. Keeping an agent's delegation chain is optional, and room in the HTTP header decides whether it survives

    Podcast · · The Non-Human & AI Identity Podcast · NHI & AI Identity Summit : The New NHI Risk Reality · Aug 26, 2026

  8. IPC passes a delegation chain between services; its signature proves only the last signer

    Draft · · IETF Datatracker · Ravi Kant Sharma, Ericsson · Jul 23, 2026

  9. A preprint cuts off every sub-agent within about 40 seconds of its parent stopping, with no revocation server

    Paper · · arXiv · Saurabh Deochake, SentinelOne · May 20, 2026

Related terms

What appears alongside AI Agent Authentication and Authorization in the same briefings.

Keel API, Inc.Credential Delegation Protocol for AI AgentsAuthZENTransaction Tokens for AgentsCross-Organizational Delegation for Workload and Agent IdentityEP Authorization Receipts for High-Risk Agent ActionsWIMSE Applicability for AI AgentsSCITT