AI Control Library / CF-04: Auditability & Evidence
Designation: Core | Control types represented: Operational
Control requirement
Plan AI audits using a documented risk-based method that prioritizes material harm, authority, data exposure, uncertainty, change and prior findings. Define the coverage and sampling needed to support each planned assurance conclusion.
Why this matters
A calendar-based review can spend equal effort on a low-impact writing assistant and an agent with payment authority. Planning should concentrate attention where a failed control would matter and identify what remains outside the review.
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
Organizational audit portfolios and individual AI engagements, including supplier dependencies, shared platforms and rapidly changing deployments.
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 audit planning owner maintains prioritization; assessment leads define engagement scope with business and technical risk input.
Implementation guidance
- Maintain an AI audit universe linked to system inventories, business criticality, autonomy, sensitive data and material dependencies.
- Prioritize using risk, uncertainty, incidents, change frequency and previous assurance rather than relying only on a model label or vendor reputation.
- Define objectives, criteria, population, sample selection, expertise and dependencies for each engagement; justify excluded risks.
- Review the plan after material events, documenting changes in scope and any resulting reduction in the assurance that can be provided.
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 |
|---|---|
| Compare priorities with material system risks | High-consequence authority and evidence gaps influence coverage. |
| Inspect a sample selection | Population, selection rationale and limits support the intended conclusion. |
| Introduce a significant incident | The plan is reconsidered rather than waiting automatically for the next annual cycle. |
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
- Audit universe and risk methodology
- Prioritization and scope decisions
- Sampling and resource plans
- Plan-change and exclusion 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 unreviewed high-impact exposure or insufficient assessment resources to the assurance sponsor. Obtain a deliberate resourcing or risk decision without representing excluded areas as assured.
Example in practice
Hypothetical example: A new tool lets an agent modify customer accounts. The audit plan brings forward access-boundary testing and adjusts a lower-impact documentation review, recording the revised coverage.
Related controls
- CF04-001: AI Audit & Assurance Policy
- CF01-002: AI Risk Management Program
- CF04-005: AI Audit Management Process
Framework mappings
- CSA AICM (A&A-03)
- NIST AI RMF (GOVERN 6.2, MANAGE 4.2)
- EU AI Act (Art. 9(7))
- ISO 42001 (Cl. 9.2)
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 Auditability & Evidence
Version 1.0