AI Control Library / CF-04: Auditability & Evidence
Designation: Core | Control types represented: Operational
Control requirement
Track AI audit findings through risk-based prioritization, accountable treatment, due dates, verification and closure. Require evidence that the underlying deficiency has been corrected or that an authorized residual-risk decision remains current.
Why this matters
Closing a ticket because a setting changed does not establish that the control now works. Remediation must address the tested failure, related exposure and the conditions under which it could recur.
How the control types work together
Operational controls establish the assigned responsibilities, review processes and response decisions described here. Their effectiveness depends on performing the work and using evidence to guide action, rather than approving a document alone.
When it applies
Findings from AI audits and assurance assessments, including recurring deficiencies and supplier or shared-control dependencies.
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 control owner implements treatment; the business risk owner accepts residual exposure and an authorized reviewer verifies closure.
Implementation guidance
- Record the criterion, failure evidence, affected scope, risk and responsible owner, preserving the original finding and management response.
- Define corrective action and interim containment with deadlines proportionate to impact and dependencies.
- Retest the relevant failure path and related configurations, checking that fixes do not introduce a new material weakness.
- Close only on supported verification or an explicit authorized risk disposition, monitoring expiry and recurring findings separately.
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 high-impact finding | Treatment, containment, owner and due date are explicit. |
| Review a closure | Retest evidence addresses the original failure on the relevant deployed version. |
| Inspect an overdue or recurring finding | Escalation and renewed risk assessment occur rather than repeated date extensions alone. |
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
- Finding and risk register
- Treatment and interim-control plans
- Retest evidence
- Closure, risk acceptance and recurrence 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 overdue critical action, failed remediation or expired acceptance to the risk authority and assurance sponsor. Restrict the affected capability where current operation exceeds accepted conditions.
Example in practice
Hypothetical example: A log-integrity finding is marked fixed after storage permissions change. Retesting also checks deletion by the administrative service account and finds the original exposure remains, so the finding stays open.
Related controls
- CF04-005: AI Audit Management Process
- CF01-004: Policy Exception Process
- CF10-010: AI Post-Incident Review & Learning
Framework mappings
- CSA AICM (A&A-06)
- NIST AI RMF (MANAGE 2.4)
- EU AI Act (Art. 9(2))
- ISO 42001 (Cl. 10.1)
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