AI Control Library / CF-05: Data Governance & Privacy
Designation: Supplemental | Control types represented: Operational and Technical
Control requirement
Restrict use of production AI data in development, testing, demonstration and evaluation to approved, necessary cases with proportionate protection. Prefer suitable lower-exposure alternatives and prevent uncontrolled replication.
Why this matters
Production information can spread into notebooks, test logs and demonstration tools that lack production safeguards. Moving data to a test environment does not remove its sensitivity or handling obligations.
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
Production records, prompts, outputs and relevant derivatives copied or accessed outside their original operational workflow.
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 exceptional use; development and evaluation owners implement minimization and environment controls.
Implementation guidance
- Define permitted nonproduction purposes and prefer representative synthetic or appropriately transformed data where sufficient.
- Require a documented necessity and scope decision for production data, including privacy, access and supplier review.
- Minimize fields and records, isolate environments and prevent unmanaged exports, telemetry or model-improvement reuse.
- Apply retention, deletion and recovery reconciliation, verifying that transformed data and derived outputs receive appropriate protection.
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 |
|---|---|
| Inspect a test dataset | Its source, purpose and authorization support the use. |
| Attempt an unapproved production export in a controlled test | The chosen restrictions prevent or detect it. |
| Close an evaluation | Copies and derived records follow the approved disposition. |
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
- Nonproduction data standard
- Necessity and approval decisions
- Environment and export tests
- Deletion and derived-data records
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 unauthorized production-data copies to privacy, security and data owners. Contain access and sharing, identify affected derivatives and apply the approved incident and disposal process.
Example in practice
Hypothetical example: A team wants real support transcripts for a public demo. It uses reviewed synthetic examples instead, keeping the demonstration’s data and telemetry outside the customer-data boundary.
Related controls
- CF05-010: AI Data Purpose Limitation
- CF05-015: Privacy-Enhancing Technologies for AI
- CF05-011: AI Data Retention & Deletion
Framework mappings
- CSA AICM (DSP-15)
- NIST AI RMF (MAP 1.1)
- ISO 42001 (Cl. 8.2)
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