AI Control Library / CF-05: Data Governance & Privacy
Designation: Core | Control types represented: Operational and Technical
Control requirement
Encrypt AI data at rest and in transit where required by classification, risk and applicable obligations, using approved maintained cryptographic services and key controls. Verify coverage across relevant stores, channels and derived copies.
Why this matters
Encryption protects specified states and boundaries, not every moment in an AI workflow. Authorized applications may still see plaintext, and logs or exports can bypass the protection applied to the primary store.
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, artifacts, retrieval resources, prompts, outputs, logs, transfers and backups according to their protection requirements.
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 defines protection needs; security sets cryptographic standards and platform teams implement coverage and key access.
Implementation guidance
- Map storage and communication boundaries, identifying where information is decrypted and which identities can use it.
- Use approved supported algorithms, protocols and services, configuring identity verification, key access and failure behavior explicitly.
- Protect intermediate stores, transfers, replicas and recovery paths; assess provider-managed boundaries and customer responsibilities.
- Test coverage, configuration drift, authentication failure and key availability, recording limitations and corrective actions.
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 |
|---|---|
| Trace a sensitive item across the workflow | Required storage and channels are protected or have an explicit treatment. |
| Exercise an untrusted endpoint or failed secure channel | The workflow follows the approved fail-safe behavior. |
| Inspect an export or backup | Protection extends to the copy and its recovery keys. |
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
- Encryption coverage and requirements
- Service and key-access configuration
- Channel and storage tests
- Drift and remediation 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
Contain sensitive paths lacking required protection and notify security and data owners. Assess exposure and restore verified safeguards without introducing uncontrolled availability failures.
Example in practice
Hypothetical example: A database is encrypted, but an AI batch export writes plaintext to a shared location. The coverage review identifies the separate boundary and protects or removes that export route.
Related controls
- CF05-S01: AI Encryption & Key Management Policy
- CF05-S05: AI Encryption Risk Management
- CF04-S01: AI Encryption Operations Monitoring & Reporting
Framework mappings
- CSA AICM (CEK-03, CEK-04)
- NIST AI RMF (MAP 1.1)
- EU AI Act (Art. 10, GDPR Art. 32)
- 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