Integration: APS data governance + DataLedger provenance standard

[aeoess]

Integration: APS data governance stack + DataLedger provenance standard

Hey @epaunova — DataLedger's approach to verifiable AI training data provenance addresses the documentation side of data attribution. We've built the enforcement side.

The Agent Passport System includes 5 data governance modules (shipped, tested, on npm):

1. Data Source Registration — cryptographic receipts for data sources with Ed25519 signatures, custodian attestation, terms composition (intersection of allowedPurposes, union of excludedPurposes)
2. Data Contribution Ledger — per-access compensation tracking, source metrics, agent footprint across all accessed sources
3. Data Enforcement Gate — enforce/audit/off modes for access control against registered source terms
4. DataGateway — terms acceptance workflow with revocation, Merkle-committed access receipts
5. Training Attribution — multi-hop derivation chains (data → summary → training), fractional contribution weights, signed training attribution receipts

Each data access produces a signed receipt with termsAtAccessTime snapshot, dataHash, and gateway signature. Receipts feed into Merkle trees for settlement. The settlement protocol generates GDPR Article 30 compliance reports automatically.

Where this maps to DataLedger: Your provenance standard defines WHAT data was used. Our enforcement gate defines WHETHER it was used compliantly. Together: verifiable provenance + cryptographic enforcement.

EU AI Act Article 10 (August 2026) requires both documentation AND governance. Neither of us alone covers both.

SDK: npm install agent-passport-system (v1.21.4, 1183 tests). The data governance modules have 120+ dedicated tests including adversarial scenarios.

[epaunova]

Thanks for the detailed mapping, the layered framing (provenance vs enforcement) makes sense. DataLedger is intentionally minimal at the provenance layer. Happy to stay in touch as both projects develop. For now keeping this issue open as a reference point for future composability discussion.