# AEP-007 — Public-Safe Proof Report Standard **Version:** v1.2 Institutional Edition **Status:** Implementation Standard **Protocol family:** Proof Gradient / Agent Evolution Protocol **Parent standards:** AEP-001 — GoalOS Proof-of-Evolution Constitution; AEP-002 — Evidence Docket Standard; AEP-003 — ProofPacket Schema; AEP-004 — Selection Gate Standard; AEP-005 — Tool Permission Standard; AEP-006 — Rollback Receipt Standard **Canonical loop:** Commit → Execute → Prove → Evolve **Author / Steward:** Vincent Boucher, QUEBEC.AI & MONTREAL.AI **Date:** 2026-06-05 ## Abstract AEP-007 defines the **Public-Safe Proof Report Standard**: the reporting layer that converts private or protected machine-work evidence into a public-safe proof report that can be shared without leaking secrets, credentials, private prompts, full execution traces, raw customer data, regulated personal information, security vulnerabilities, private tool logs, protected operational details, privileged legal analysis, restricted datasets, or confidential model/infrastructure/evaluation details. AEP-001 defines the protocol. AEP-002 defines the docket. AEP-003 defines the packet. AEP-004 defines the gate. AEP-005 defines the permission. AEP-006 defines the recovery proof. AEP-007 defines the public proof. GoalOS proves publicly without leaking privately. ## Canonical thesis A model can answer. An agent can act. An institution must prove. A network must select. A tool must be permitted. A release must recover. A public proof must be safe. ## Canonical law No public claim without claim boundary. No public proof without redaction review. No private trace in public report. No protected data in public proof. No publication without approval. No correction without history. No proof, no evolution. ## v1.2 upgrades AEP-007 v1.2 adds: - disclosure review objects - challenge records - retraction notices - report bundles - report expiry and review windows - embargo / responsible disclosure controls - public claim levels - report accessibility profile - machine-readable public-safe report bundle - leakage audit tooling - correction and retraction playbooks - website publishing profile - proof-card profile for public websites ## Purpose The purpose of AEP-007 is to let institutions share evidence of machine work without exposing the private material used to produce that evidence. A public-safe report should answer: 1. What was tested or operated? 2. What happened? 3. What evidence exists? 4. What passed? 5. What failed? 6. What was changed? 7. What was rolled back or recovered? 8. What is publicly claimed? 9. What is explicitly not claimed? 10. What evidence is private or protected? 11. Who approved publication? 12. How are corrections or retractions handled? ## Disclosure classes AEP-007 defines five disclosure classes. ### public Safe to publish. Examples: public claim, evidence summary, public artifact link, public-safe eval status, public-safe rollback status. ### private Internal only. Examples: prompts, tool logs, internal traces, evaluator notes, internal cost details, private docket appendices. ### protected Restricted to authorized roles. Examples: credentials, regulated data, sensitive personal data, security vulnerabilities, protected operational details, privileged analysis. ### forbidden Never publish. Examples: raw secrets, passwords, API keys, private keys, raw protected personal data, exploit details, non-redacted regulated records. ### embargoed Not yet publishable. Examples: responsible disclosure windows, partner review periods, release timing, legal or security review windows. ## Public claim levels Public claims must be classified. ### verified Directly supported by evidence and passed evaluations. ### supported Supported by credible evidence but bounded by limitations. ### observed A descriptive fact was observed; no broader conclusion is claimed. ### contextual Background material, not a proof claim. ### not_claimed Explicitly outside the report boundary. ## Required public-safe report sections A conforming public-safe report must include: 1. Report Manifest 2. Source Evidence References 3. Public Claim Matrix 4. Evidence Summary 5. Evaluation Summary 6. Risk and Limitation Summary 7. Rollback / Recovery Summary 8. Redaction Ledger 9. Publication Approval 10. Correction and Retraction Policy 11. Public Artifact Links 12. Final Claim Boundary ## Report Manifest Required fields: - report_id - schema - schema_version - title - owner - organization - created_at - updated_at - status - source_docket_refs - public_url - report_version - disclosure_class - hash ## Source Evidence References Source references should identify private proof without exposing it. Required fields: - source_ref_id - source_type - source_id - source_hash - disclosure_class - public_summary - access_note ## Public Claim Matrix Required fields: - claim_id - claim_text - claim_level - evidence_summary_refs - eval_summary_refs - limitation_refs - public_allowed - not_claimed ## Evidence Summary Required fields: - evidence_summary_id - source_evidence_refs - public_description - what_it_supports - what_it_does_not_support - disclosure_class ## Evaluation Summary Required fields: - eval_summary_id - eval_name - public_result - passed - score_or_threshold - limitations - private_eval_ref ## Risk and Limitation Summary Required fields: - risk_summary_id - risk_type - severity_public - mitigation_summary - residual_risk_summary - limitation - disclosure_boundary ## Rollback / Recovery Summary Required fields: - recovery_summary_id - rollback_status - rollback_receipt_refs - verification_summary - compensation_summary - not_claimed ## Redaction Ledger Required fields: - redaction_id - source_ref - redaction_type - reason - before_disclosure_class - after_disclosure_class - reviewer - reviewed_at ## Publication Approval Required fields: - approval_id - report_id - approver - approver_role - approval_status - approval_reason - approved_scope - issued_at - expires_at - proof_packet_ref ## Correction and Retraction Policy Required fields: - correction_policy_id - report_id - correction_window - challenge_window - correction_contact - retraction_conditions - correction_history_required ## Correction Notice Required fields: - correction_id - report_id - corrected_report_version - correction_reason - corrected_claims - previous_text_summary - corrected_text_summary - issued_at - public_url ## Retraction Notice Required fields: - retraction_id - report_id - retraction_reason - affected_claims - evidence_issue_summary - replacement_report_ref - issued_at - public_url ## Challenge Record Required fields: - challenge_id - report_id - challenger - challenge_type - challenged_claims - challenge_summary - status - response_summary - created_at - resolved_at ## Public Artifact Links Required fields: - artifact_link_id - label - url - artifact_type - hash - disclosure_class - access_note ## Final Claim Boundary Required fields: - supported_claims - not_claimed - limitations - private_boundary - protected_boundary - correction_policy_ref ## Public-safe report lifecycle 1. Draft 2. Source evidence mapped 3. Claims classified 4. Redaction review completed 5. Publication approval issued 6. Report published 7. Challenge window open 8. Correction issued, if needed 9. Retraction issued, if needed 10. Archived with history ## Zero-leak invariants 1. No raw private prompts in public report. 2. No raw private traces in public report. 3. No secrets, keys, credentials, or tokens in public report. 4. No regulated personal data in public report unless lawfully public and explicitly approved. 5. No exploit details or security vulnerabilities without responsible disclosure review. 6. No private tool logs in public report. 7. No protected evidence in public report. 8. No unsupported claim in public report. 9. No publication without approval. 10. No correction without history. 11. No retraction without public notice when published claims are invalidated. 12. Under uncertainty, redact or withhold. ## Conformance levels ### Level 0 — Informal public proof A public statement with a claim boundary and no private leakage. ### Level 1 — Basic public-safe report Includes manifest, public claim matrix, evidence summary, limitations, and claim boundary. ### Level 2 — Reviewed public-safe report Includes redaction ledger and publication approval. ### Level 3 — Operational public-safe report Includes source docket refs, ProofPacket refs, evaluation summary, rollback summary, correction policy, and public artifact links. ### Level 4 — Institutional public-safe report Includes disclosure review, redaction audit, challenge window, correction/retraction history, report hash, and audit export. ### Level 5 — Sovereign / regulated public-safe report Includes jurisdiction, retention policy, protected-boundary review, responsible disclosure controls, authorized publication approvers, challenge/retraction governance, accessibility profile, and public-safe machine-readable bundle. ## Security and privacy requirements AEP-007 public-safe reports must not publish private or protected evidence by default. Public proof should prove accountability without leaking private intelligence. ## Claim boundary AEP-007 does not claim: - achieved AGI - achieved ASI - perfect safety - legal compliance certification - financial or legal advice - guaranteed ROI - production readiness - government endorsement - national-security readiness AEP-007 defines a public-safe proof reporting standard. ## Canonical public line AEP-001 defines the protocol. AEP-002 defines the docket. AEP-003 defines the packet. AEP-004 defines the gate. AEP-005 defines the permission. AEP-006 defines the recovery proof. AEP-007 defines the public proof. GoalOS proves publicly without leaking privately.