Hardware-rooted attestation for AI-agent evidence: composing IETF RATS with action evidence packages

An action evidence package (AEP) is a signed, append-only record of what an AI agent did, who or what authorised the action, and what the outcome was. It is a software-layer artefact: it tells a verifier the story of an action as the agent's own runtime reports it. This note argues that software attestation of this kind is necessary but not sufficient. When a verifier's question shifts from 'what does the agent claim it did?' to 'did the specific model version the operator claims to have deployed actually produce this output, on unmodified hardware?', the AEP alone cannot answer. The missing element is a hardware root of trust. The IETF Remote Attestation Procedures (RATS) architecture (RFC 9334) and Veraison, an open-source RATS Verifier implementation (Confidential Computing Consortium / Linux Foundation), supply exactly this. We propose a composite attestation -- hardware Evidence appraised under RATS, bound to a software AEP -- and map a small verifier vocabulary (Authorised / Unauthorised / Indeterminate / Attested / Contested / Expired) onto RATS appraisal outcomes. A feasibility experiment using swtpm 0.7.3 demonstrates the binding and three-way platform verdict end to end: Attested for a good, fresh quote; Contested when the model measurement is swapped; Expired when a stale quote is replayed; and rejection of a forged AEP outcome bound to a valid quote.

Note. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible.

Version 2 (2026-07-17): removed leftover editorial placeholders ([VERIFY: ...] markers) from the reference list and the stale target-venue line; body text unchanged from version 1.