Platform

Execution assurance

Five verbs get collapsed into one word in most agent-security marketing. They are not the same guarantee, and the difference is where audits bite.

StageQuestion it answersWho is trustedAWARE today
AuthoriseIs this agent permitted to attempt this at all?AWAREimplemented policy engine decides
ApproveHas a human signed off on this specific instance?The approverimplemented approval requirement on tools
RevalidateIs the authority still valid at the moment of action, after any state change?AWAREpartial session and revocation checks, no general revalidation hook
CommitDid the change land atomically, or half?The target systemroadmap Commit Guard, not implemented
VerifyDid the target actually do what was authorised?An independent partyroadmap no execution receipt
The honest statement AWARE today authorises and records. It does not verify that an authorised action was carried out as intended, because nothing in the current design observes the target after the fact. A Commit Guard that makes consequential changes atomic, target-side enforcement that makes the target verify AWARE’s authorisation, and independent execution verification that produces a receipt — all three are specified in the roadmap architecture and none of them is implemented today.

Why we publish this page

A buyer who believes AWARE already verifies execution, and discovers otherwise during an audit, loses the account and the credibility. This page exists so that nobody has to discover it later.