Decision Log
The public record of AI decisions made against CertifiedData certified artifacts. Each entry is cryptographically signed, appended to an append-only log, and independently verifiable without an account.
This page shows decisions logged with public_mode=true. Private records are not published. Machine-readable: GET /api/decision-log
How decision records reach this log
Artifact certified
A dataset or AI artifact is cryptographically certified and registered in the artifact registry.
Artifact used
A model or workflow consumes the certified artifact to produce a decision or output.
Decision logged
The operator logs the decision via POST /v1/decisions, optionally referencing the certificate ID.
Published here
Records with public_mode=true are Ed25519-signed, chain-linked, and appear in this log.
Live public records· showing 1–25 of 8,302
Record structure and API
Public API endpoints
GET /api/decision-log
?limit=25&offset=0
GET /api/decision-log/latest
GET /api/decision-log/:id
GET /api/decision-log/:id/verify
GET /.well-known/decision-signing-keys.json
Response fields (public records)
decision_id — unique identifier
decision_label — event type
selected_option — outcome
certificate_id — linked artifact (optional)
public_hash — SHA-256 of this record
previous_public_hash — chain link
signature_alg — Ed25519
integrity_status — verified | pending
timestamp — ISO 8601
Integrity model
Each public record is signed with Ed25519 using the key identified in /.well-known/decision-signing-keys.json.
The previous_public_hash field links each record to the prior entry. Any modification to a prior record changes its hash and breaks the chain.
Signature verification is available at /api/decision-log/:id/verify without authentication.
Current implementation scope
Public decision records are available via the API and on this page. Records are optionally linked to certified artifact IDs.
Ed25519 signing and SHA-256 chain linking are active for all public records. The /verify endpoint performs live signature checks.
Only records explicitly submitted with public_mode=true appear here. Private records are not published.
Log a decision
Use POST /v1/decisions with an authenticated API key. Set public_mode: true to publish the record to this log. Reference a certificate_id to link the decision to a certified artifact.
POST /v1/decisions
Authorization: Bearer <api_key>
{
"decision_label": "loan_approved",
"selected_option": "approved",
"public_mode": true,
"artifact_reference": {
"certificate_id": "cert_01j9k..."
}
}Questions
What is a decision record?
A decision record is a structured log entry that captures an AI system action: the decision label, selected outcome, input context, model version, policy version, and timestamp. Records submitted with public_mode=true are signed with Ed25519, assigned a SHA-256 hash, and appended to this public log.
How is the decision log different from the artifact registry?
The artifact registry records certified datasets and AI artifacts — what was built and signed. The decision log records how those artifacts were used downstream — decisions made by systems that consumed certified data. Certification is about provenance; decision logging is about use.
How does decision lineage connect to artifact certification?
Each decision record can reference a certificate_id, linking the decision to the specific certified artifact used. This creates a traceable chain: certified dataset → model trained on it → decision made by that model. The chain allows auditors to follow a decision backwards to its data origin.
Can decision records be independently verified?
Each record includes an Ed25519 signature and a chain hash linking it to the prior entry. The signing public key is available at /.well-known/decision-signing-keys.json. Any party with the record payload and the public key can verify the signature without contacting CertifiedData.
Are all AI system decisions published here?
No. Only decisions logged via POST /v1/decisions with public_mode=true appear in this log. Private decision records are retained but not published. Operators choose what to make public at the record level.
Why does decision logging matter for governance?
AI governance requires verifiable evidence of how systems behave — not just what data they were trained on. Decision logs provide timestamped, cryptographically signed records of individual AI actions, satisfying audit and traceability requirements for frameworks like the EU AI Act (Articles 12 and 19).
Related
AI Decision Logging
What decision logging is, how it supports EU AI Act compliance, and how it differs from general audit logs.
AI Decision Lineage
The traceable chain from certified dataset to model to downstream decisions — and why it matters for governance.
Artifact Registry
The public index of certified AI artifacts. Decision records can reference any registry entry by certificate ID.
Transparency Log
Browse the append-only public audit trail for certificates, datasets, and decision records in one place.
Verify a Certificate
Independently verify any artifact certificate using its SHA-256 fingerprint and Ed25519 signature.
Verification Docs
How to programmatically verify certificates and decision records using standard cryptography libraries.
Synthetic Data Certification
Certify the synthetic datasets referenced in decision records. Each certificate ID links a decision to its verified artifact.
Payment Decision Log
AI agent spend authorization decisions — every payment authorized, blocked, or escalated, with policy hash and reason codes.