Operational records for
AI governance.
Keep preliminary assessments, operational controls, and metadata-only evidence attached to the AI system and its real implementations.
Read the governance docsPinned catalog · reviewer rationale retained
SHA-256 6f29c3...f62aImmutable · metadata only
- Human oversightconfigured
- Disclosure verificationneeds evidence
- Incident proceduregap
Govern the AI system across its implementations
An AI system is a project-scoped record independent of any single agent. Link the environments, deployments, and agents that implement it so evidence follows the real operating surface as it changes.
- 1
Register
Describe the use case, owner, geographies, affected persons, and implementation links.
- 2
Assess
Record an immutable, versioned answer set against a pinned catalog version.
- 3
Review
Approve with rationale, a supported provider or deployer role, and uncertainty acknowledged.
- 4
Operate
Work generated controls, transparency records, monitoring plans, and incidents.
- 5
Export
Capture a metadata-only snapshot and stream its hash-verifiable archive.
Turn the latest approved assessment into work
Controls retain the governing article, applicability rationale, accountable implementation type, status, evidence state, and change history.
Applicable control recorded without an implementation.
gapThe control is marked as needing supporting evidence; evidence status is tracked separately.
needs_evidenceImplementation recorded; evidence links are tracked separately.
configuredException recorded with its required rationale.
not_applicable
Connic · Customer · Shared · External
Refreshing re-derives controls from the current latest approved assessment while preserving prior changes.
Track disclosure policy and application delivery separately
The governance record captures what your organization says it implements and the evidence that supports that statement.
- Disclosure copy, locales, placement, frequency, and display mode
- Customer implementation and upstream-provider evidence
- Attested applications and supporting verification
- User-facing disclosures and synthetic-content marking
- Any prompt, response, stream, or connector-traffic modification
- Prefixes, banners, metadata, or media watermarks
AI Governance supports documentation and evidence workflows. It does not provide legal advice, make legal determinations, issue certifications, or replace professional counsel.
Export an immutable, hash-verifiable evidence archive
A snapshot is an immutable, metadata-only capture scoped to selected systems and environments, with an optional time window.
{
"snapshot_id": "8f2d3e43-90ab-4f7b-a218-687719b0c6e3",
"content_sha256": "6f29c3e8a1095cbd1bd9f645cb7b355b16837e97b1dbca38344752103fddf62a",
"scope": {
"system_id": "5403e168-b9fd-4e42-8556-8a68d97d899c",
"environment_ids": ["1cb302be-64d7-4c08-9d25-337a6897cb82"],
"telemetry_scope": {
"runs": "system_agent_environment_pairs",
"approvals": "system_agent_environment_pairs",
"judges": "system_agent_environment_pairs"
},
"time_from": "2026-07-01T00:00:00Z",
"time_to": "2026-07-31T23:59:59Z",
"metadata_only": true
},
"record_counts": {
"systems": 1,
"assessments": 1,
"controls": 3,
"transparency_policies": 1,
"transparency_applications": 1,
"monitoring_plans": 1,
"incidents": 0
},
"telemetry_file": "telemetry-summary.json",
"raw_prompt_or_output_content_included": false
}Immutable
Snapshots cannot be edited or deleted and persist for the project lifetime.
Content-addressed
The canonical payload and every data or report artifact listed in the manifest carry SHA-256 hashes.
Portable
The ZIP includes a human report, canonical JSON, collection CSV and JSON, and a manifest.
Scope-aware
Coverage records selected environments, time range, retention, and missing evidence.
Privacy-bounded
Summary telemetry is included without copying raw prompt or output content into the archive.
Hashes detect changed content. They are not signatures, do not establish who published an archive, and do not certify legal sufficiency. Review the evidence and governance boundaries.
Keep exploring
Approvals
Preserve human decisions around sensitive actions.
Observability
Attach operating telemetry to the governed system.
Judges
Carry evaluation evidence into the record.
Guardrails
Enforce documented runtime boundaries.
Environments
Scope systems and exports to real environments.
REST API
Automate the complete governance workflow.