assurance

This paper extends the Auditable Agents framework's Policy Checkability dimension to claims about AI agents, defining when a claim is auditable and proving that independent record coverage, per-action authorization binding, and completeness are necessary for verification, with practical guidance for operators and standard setters.