Guidelines  ·  2026-11-18

IETF draft-stafordtituss-agentic-mandate-evidence-00: 'Continuity Evaluation of Agent Delegation Records'

GuidelinesMedium impactGlobal
New IETF individual Internet-Draft, draft-stafordtituss-agentic-mandate-evidence-00, 'Continuity Evaluation of Agent Delegation Records' (last updated 2026-11-16; dated 17 November 2026; Informational). It defines a deterministic, three-valued evaluation (preserved / violated / unresolved) over AUDIT intent/delegation/authorization/action records to answer whether the executed effect stayed within delegated authority, plus a claim model, permitted signer roles, a mapping onto AUDIT record types, and interoperable test vectors. Derived from the Agentic Mandate Evidence Protocol (AMEP).
Agentic action happens through chains of delegation (user → agent → sub-agents → tools) where each hop can widen what is done relative to what was asked. This draft provides a concrete, machine-checkable continuity evaluation to detect reuse of an approval for another command, a sub-agent with broader scope than its parent, or untrusted content steering a mutation — directly supporting auditability and accountability for autonomous agents.
Review the draft and the AUDIT/AMEP record model for applicability to your agent-authorization evidence; consider contributing feedback/comments to the AUDIT BOF and test-vector work; feed the three-valued evaluation into agent audit trail and accountability design.
draft-stafordtituss-agentic-mandate-evidence-00
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →