CAP v1.0 — Official Release

CAP v1.0 - Official Release

The first official release of Content / Creative AI Profile (CAP), the world's first open specification for cryptographic AI content refusal logging.

✨ Highlights

🔐 Safe Refusal Provenance (SRP)

• Cryptographic proof of what AI refused to generate
• GEN_ATTEMPT → GEN_DENY event chain

📊 Completeness Invariant

∑ GEN_ATTEMPT = ∑ GEN + ∑ GEN_DENY + ∑ GEN_ERROR

Mathematical guarantee that all attempts have recorded outcomes

🏅 Three Conformance Levels

Level	Requirements
Bronze	Hash chain, basic logging, 6-month retention
Silver	+ SRP, external anchoring (daily), 2-year retention
Gold	+ Real-time verification, HSM, 5-year retention

⚓ External Anchoring

• RFC 3161 TSA
• SCITT Transparency Services
• Blockchain anchoring

📋 Regulatory Alignment

• EU AI Act Article 12
• Digital Services Act Article 37
• Colorado AI Act (SB24-205)
• TAKE IT DOWN Act

📚 Documentation

• Full Specification
• Academic Paper (DOI: 10.5281/zenodo.18213616)
• World-First Verification Report

🔗 Related

• VAP Framework v1.2 — Parent framework
• VCP v1.1 — Finance profile
• Reference Implementation

License

CC BY 4.0 International

---

"Verify, Don't Trust"

CAP v0.2.0 — World’s First Safe Refusal Provenance Specification

This release publishes CAP (Content / Creative AI Profile) v0.2.0, the first open specification to enable cryptographically verifiable proof that AI content generation requests were refused.

CAP v0.2.0 introduces Safe Refusal Provenance (SRP) to address a structural gap exposed by recent AI content incidents: while platforms can claim safeguards exist, they have lacked any externally verifiable way to prove that disallowed content was not generated.

Key highlights

• Safe Refusal Provenance (SRP)

  • Cryptographically records GEN_ATTEMPT and refusal outcomes (GEN_DENY)
  • Enables third-party verification of non-generation events

• Completeness Invariant

  • Every generation attempt MUST have exactly one verifiable outcome
  • Prevents selective logging, omission attacks, and post-hoc narrative rewriting

• Integrated AI Lifecycle Audit

  • Unified event model across INGEST → TRAIN → GEN → EXPORT
  • Covers rights basis, consent state, model context, and delivery

• Regulator-ready Evidence Pack

  • Structured audit artifacts aligned with:
    • EU AI Act Article 12 / Article 53
    • Digital Services Act Article 35
    • GDPR (including crypto-shredding design)
    • NCII-related obligations (e.g., TAKE IT DOWN Act)

• Open, auditable specification

  • Normative Markdown specification
  • JSON Schemas for machine validation
  • Examples and conformance test vectors included

World-first verification

Independent consolidated research confirms that CAP-SRP is the world’s first open specification to:

• Cryptographically prove AI refusals (non-generation)
• Enforce a completeness invariant for AI moderation events
• Provide regulator- and au