AI Control Library / CF-04: Auditability & Evidence
Designation: Supplemental | Control types represented: Operational and Technical
Control requirement
Record security-relevant cryptographic key lifecycle actions for AI services, including creation, authorization, activation, rotation, suspension, recovery and destruction, with attributable identities and protected evidence.
Why this matters
Encrypted data can become exposed or unavailable through unauthorized key actions. Investigation needs to establish what changed and who authorized it without collecting the secret key material itself.
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
Keys protecting AI data, model artifacts, logs, backups and integrations, including externally managed keys where relevant audit evidence can be obtained.
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 key-management owner defines events and access; security monitors activity and data owners assess availability or disclosure impact.
Implementation guidance
- Define the lifecycle events and sensitive key-use events needed for the threat model, recording identifiers, version, actor, authorization, time and result.
- Enable protected audit trails across key services and dependent platforms, explicitly excluding secret key values and recovery material.
- Link key changes to approvals and affected services, distinguishing scheduled rotation from unusual disablement, access expansion or export attempts.
- Test event capture, privileged access and incident retrieval, including failed requests and provider visibility limitations.
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 |
|---|---|
| Perform an authorized test rotation | Old and new key references, actor, approval and result are attributable. |
| Attempt an unauthorized key action in a safe test | The denial and relevant context are available for review. |
| Inspect captured records | They support investigation without revealing keys or recovery secrets. |
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
- Key audit event specification
- Audit service configuration
- Lifecycle and authorization records
- Capture, access and retrieval tests
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 unexplained key actions or missing audit trails to security and key-management owners. Restrict compromised authority, assess protected data and coordinate revocation or recovery to avoid unmanaged service loss.
Example in practice
Hypothetical example: A key is disabled outside the maintenance window. The audit trail identifies the requesting identity and failed downstream decryptions, allowing response teams to distinguish an authorized recovery activity from misuse.
Related controls
- CF04-010: AI Audit Log Protection & Integrity
- CF05-S12: AI Encryption Key Lifecycle Management (Activation/Suspension/Deactivation)
- CF05-S16: AI Encryption Key Inventory Management
Framework mappings
- CSA AICM (LOG-11)
- NIST AI RMF (MANAGE 4.1)
- ISO 42001 (Cl. 9.1)
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 Auditability & Evidence
Version 1.0