Guidelines  ·  2026-11-20

IETF draft-cosme-audit-agent-traceability-00: 'Delegation and Interaction Traceability for AI Agents: An Attestation and Registry Evidence Model'

GuidelinesMedium impactGlobal
New individual Internet-Draft (last updated 2026-11-18), draft-cosme-audit-agent-traceability-00, describing requirements and an evidence model for delegation and interaction traceability across AI-agent chains. It is built on an append-only, hash-chained exhibit registry (each action recorded exactly once, counters reconciled) plus per-action attestation published at kernel level with measured sub-microsecond publication latencies. It defines a traceability record, states what a verifier can and cannot conclude, identifies where existing agent identity/authorization/transport work needs a common traceability layer, and (notably) defines no wire protocol of its own — it sets requirements against which formats defined elsewhere can be evaluated.
Poses concrete, measurable requirements (hash-chained attestation registry, kernel-level publication latency) for establishing that a given AI agent action happened and who delegated it. Useful as an evaluation rubric for the other agent-audit/evidence protocols that are proliferating, and it highlights that delegation chains are the key gap in current agent security work.
Evaluate agent-audit solutions against the stated requirements; track for convergence with identity/authorization drafts.
IETF Datatracker — draft-cosme-audit-agent-traceability-00
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →