AI Control Library / CF-05: Data Governance & Privacy
Designation: Core | Control types represented: Operational and Technical
Control requirement
Define and implement retention and deletion schedules for AI information and relevant derivatives, coordinating suppliers, indexes, caches, logs and backups. Verify results and document justified exceptions and practical limits.
Why this matters
Data often survives beyond its source through derived stores and operational copies. A deletion promise must distinguish completed removal from restrictions, scheduled expiry and effects that cannot be established.
How the control types work together
Operational controls define scope, authority and the actions required when conditions are not met. Technical controls produce evidence or enforce the relevant boundaries. Both must apply to the same deployment and operating conditions.
When it applies
AI datasets, prompts, outputs, retrieval indexes, embeddings, logs, evaluation material and recovery copies within the assessed processing scope.
Define the assessed boundary, dependencies and evidence limitations. Tailor depth and review frequency to potential impact, and document why a particular safeguard or exclusion is appropriate to the actual deployment.
Ownership
The data owner approves retention; stewards and platform teams execute deletion with privacy and records-management review.
Implementation guidance
- Assign retention periods and triggering events by purpose, sensitivity and applicable requirements, avoiding indefinite defaults.
- Map deletion dependencies across active stores, derivatives and suppliers, separately addressing backups and relevant model training uses.
- Automate expiry where appropriate, handle authorized preservation requirements and prevent deleted information from reappearing through routine recovery.
- Test removal and access restrictions, record exceptions and report the actual scope of completion without assuming weights or unknown copies were erased.
How to verify it
Use representative records and authorized controlled tests. Identify the relevant version, operating conditions and expected result before testing, and retain enough evidence to support the conclusion.
| Test | Expected result |
|---|---|
| Expire a representative test item | Relevant active copies and indexes are removed or treated according to the approved plan. |
| Restore a backup in isolation | Required deletions are reconciled before returning recovered information to use. |
| Review a supplier confirmation | It covers the relevant stores and identifies limitations. |
Record a finding when required results cannot be demonstrated or practice differs materially from the approved conditions. Missing evidence limits the conclusion; it should not be treated as a successful test.
Evidence to retain
- Retention schedule and scope
- Deletion workflows and results
- Preservation and exception decisions
- Restore and supplier reconciliation
Protect evidence according to its sensitivity, access needs and retention rules. Avoid unnecessary copies of personal information, secrets or confidential material, while preserving the records needed for authorized review.
If the control fails
Escalate failed deletion or unjustified continued retention to data and privacy owners. Restrict access or use while the affected copies and obligations are resolved.
Example in practice
Hypothetical example: A document is deleted from storage but remains searchable in a vector index. Verification catches the incomplete workflow and adds index and cache handling before completion is reported.
Related controls
- CF05-002: AI Data Inventory
- CF05-009: AI Data Subject Rights Implementation
- CF05-016: AI Data Secure Disposal
Framework mappings
- CSA AICM (DSP-16)
- NIST AI RMF (MANAGE 4.1)
- EU AI Act (GDPR Art. 5(1)(e), EU AI Act Art. 12)
- ISO 42001 (Cl. 8.2, A7.1)
Review and maintenance
Review at the interval approved by the control owner and after relevant changes, incidents or failed assessments. Update affected procedures and configuration, address evidence gaps and verify corrective action before closing findings.
Back to Data Governance & Privacy
Version 1.0