AI Control Library / CF-04: Auditability & Evidence
Designation: Core | Control types represented: Operational and Technical
Control requirement
Restrict access to AI audit records and logging administration according to assigned duties, with attributable identities, approval, periodic review and accountability for searches, exports and changes.
Why this matters
Logs can contain sensitive prompts, customer information and investigation detail. Broad read access creates disclosure risk, while broad administrative access may undermine the evidence used to hold operators accountable.
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
Log viewers, search services, archives, exports, evidence repositories and logging management interfaces, including service identities and supplier support access.
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 log data owner approves access; identity and platform teams enforce it and security reviews privileged activity.
Implementation guidance
- Define roles for operational review, investigation, evidence export and logging administration, separating content access and configuration authority where practical.
- Grant least-privilege access through attributable identities, limiting tenant, dataset, time range and sensitive fields as required.
- Record access, exports and privileged changes in a protected audit path, with controlled emergency access and subsequent review.
- Review access after role changes and at approved intervals; remove obsolete rights and test service-account and shared-platform boundaries.
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 |
|---|---|
| Test an unauthorized user or tenant query | Protected records are not returned. |
| Trace an export or privileged change | An attributable identity, authorization and activity record are available. |
| Inspect a departed or reassigned operator | Access is removed or adjusted without retaining unnecessary investigation privileges. |
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
- Log access matrix and approvals
- Identity and query restrictions
- Access and export audit records
- Recertification and revocation 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 unauthorized disclosure, unattributed exports or excessive privilege to security and the log data owner. Revoke or narrow access and preserve records needed to investigate exposure.
Example in practice
Hypothetical example: A help-desk dashboard exposes full customer prompts to all support staff. Role-based views separate routine operational metrics from content that only authorized investigators need to inspect.
Related controls
- CF04-009: AI Input & Output Logging
- CF04-010: AI Audit Log Protection & Integrity
- CF12-002: AI System Least Privilege Access Control
Framework mappings
- CSA AICM (LOG-04)
- NIST AI RMF (MANAGE 4.1)
- EU AI Act (Art. 12)
- 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