HJS: Accountability Receipts for AI Agents A Minimal JEP Profile for Exportable AI Receipts

This document defines HJS, a minimal accountability receipt infrastructure for AI agents. HJS is a profile of the Judgment Event Protocol (JEP) and does not define an independent event signing protocol. HJS uses JEP events to bind signed event claims to AI-agent behavior records, receipt manifests, optional privacy-preserving human participant references, and optional deployment-specific evidence references. The HJS core is intentionally small. It defines behavior-record digest binding, receipt manifests, receipt bundles, and validation of cryptographic and structural consistency. All other capabilities, including human privacy modes, participant-supplied references, post- event review references, explanation-material references, risk descriptors, model evidence, tool-call evidence, policy-check evidence, multi-party export, and cryptographic capability profiles, are optional extensions or deployment profiles. HJS is infrastructure. It does not assign legal liability, prove subjective intent, define governance rules, enforce monitoring, define fairness, define appeal rights, define explanation rights, determine authorization validity, or establish regulatory compliance.