Guidelines  ·  2026-11-18

IETF draft-gunuganti-agent-approval-evidence-00: 'Owner-Approved Agent Spending with Sealed Evidence'

GuidelinesMedium impactGlobal
New IETF individual Internet-Draft, draft-gunuganti-agent-approval-evidence-00, 'Owner-Approved Agent Spending with Sealed Evidence: The Approval-Evidence Object for Agent Action Protocols' (last updated 2026-11-15, Informational). It describes a running system that removes the trust step for agent approvals — an owner-set, machine-checkable mandate; a deterministic check performed and recorded before any spendable token exists; owner approval on a surface the agent cannot operate/imitate; a single-use decision token bound to the hash of the exact action approved; and a sealed, owner-held receipt that grades evidence (approval PROVEN; performer completion report REPORTED/PINNED). It also specifies an optional approval-evidence object that agent action protocols can carry so a relying party can verify an approval mechanically.
Today's agent protocols (e.g., Personal Agent Protocol Operations extension) leave the approval as an assertion by the agent, 'working on trust'. This draft defines the evidence and receipt format needed for a relying party (merchant, issuer, arbitrator, principal) to verify what the owner actually approved and what happened next — directly addressing loss allocation and non-repudiation for agent-initiated transactions, a live issue given recent frontier-lab payment wallets.
Review the approval-evidence object and the mandate/sealed-receipt model for agent-payment and agent-action protocols; use it to inform fraud/dispute evidence-retention and to support machine-checkable approval in agent commerce.
draft-gunuganti-agent-approval-evidence-00
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →