Back to Gap Map Product / UX / Agent identity & authorization

AI Identity Management System (WIMSE)

Internet Engineering Task Force
source available / Overall score: 1.0

AI Identity Management System (AIMS) is an IETF WIMSE working group draft that sets out how to authenticate and authorize AI agents by composing existing standards rather than defining new protocols. Each agent gets a WIMSE or SPIFFE workload identifier with cryptographically bound credentials, OAuth 2.0 carries delegated authority, and the draft covers human approval, transaction tokens and tamper-evident audit records.

Read against revision 01. An Internet-Draft is a work in progress that is revised, replaced or left to expire, so a later revision may change what it defines.

Openness

2 medium confidence
2.0
license
IETF-Trust-Legal-Provisions (the draft text under BCP 78 and TLP 5: copy, distribute and translate it unmodified; no license to modify it outside the IETF Standards Process; not on the software tier list)+BSD-3-Clause (Code Components extracted from the draft, under the Revised BSD License, TLP 5 section 4)
source
public (revision 01 in the IETF draft archive, edited in ietf-wg-wimse/draft-ietf-wimse-aims)
core features withheld
no — an IETF working group document on the IETF stream; the working group holds the text, and no vendor sells an edition of it

The draft is published in full, anyone may copy and distribute it unmodified, and code extracted from it comes under the Revised BSD License. The IETF Trust grants no license to modify the text outside the IETF standards process, so the document itself cannot be reworked the way an openly licensed specification can.

Adoption

1 low confidence
1.0

The draft has no package of its own, so GitHub stars on the working group's editing repository are the only signal. A star is not a use, and deployments that follow the practices it describes are not counted.

Capability

not assessed

AIMS is an informational draft that, in its own words, describes how existing standards can be applied rather than defining new protocols: agent identity comes from WIMSE identifiers, delegation from OAuth token exchange and transaction tokens, approval from CIBA and decisions optionally from AuthZEN. It defines none of those mechanisms itself, so there is nothing of its own to place beside the specifications and products it draws on.

  • https://datatracker.ietf.org/doc/draft-ietf-wimse-aims/ recorded 2026-11-20

    draft-ietf-wimse-aims-01, "Intended status: Informational"; abstract: "Rather than defining new protocols, this document describes how existing and widely deployed standards can be applied".

  • https://www.ietf.org/archive/id/draft-ietf-wimse-aims-01.txt recorded 2026-11-20

    "An agent participating in this framework MUST be assigned exactly one WIMSE identifier"; these identifiers "MUST be used by resource servers ... to determine if access to a resource should be allowed"; transaction tokens yield "a downscoped token that is bound to a specific transaction".

  • https://www.ietf.org/archive/id/draft-ietf-wimse-aims-01.txt recorded 2026-11-20

    Section 10.8: an agent "can use the OpenID Client Initiated Backchannel Authentication (CIBA) protocol" so the user can "approve or deny the requested operation"; section 11: "Audit records MUST be tamper-evident".

Verified 2026-11-20, except some scores, still awaiting re-confirmation