AI Control Library / CF-01: Lifecycle Governance
Designation: Core | Control types represented: Operational
A risk assessment loses value if no one acts on its findings. AI risks also change with data, permissions, models and deployment context, so an initial score cannot serve as permanent approval.
Control requirement
Operate an AI risk management program that identifies, assesses, treats and monitors risks throughout each system’s lifecycle. Assign every material risk an owner, treatment decision, due date and residual-risk acceptance authority.
Why this matters
The control is effective when its required decisions and safeguards can be demonstrated in the actual workflow. An approved document or completed form alone is insufficient if the described actions are not performed.
How the control types work together
Operational controls provide the policies, decisions and recurring processes described here.
When it applies
AI proposals, existing systems, material changes and retirement activities; depth should reflect potential impact and uncertainty.
Tailor depth and review frequency to potential impact, system authority and uncertainty. Document the reason for any exclusion; a supplemental designation does not make an otherwise applicable safeguard optional.
Ownership
The AI risk lead owns the method; system owners maintain records and authorized business risk owners accept residual exposure.
Implementation guidance
- Define risk criteria covering people, security, privacy, reliability and business impact; record uncertainty and assumptions alongside ratings.
- Assess the complete workflow, including vendors, data flows, human review and agent tools; identify affected groups and failure scenarios.
- Select treatment, ownership and deadlines for each material risk; distinguish acceptance from remediation and require the proper authority.
- Review risk after incidents, material changes and scheduled reassessment; verify that completed treatments actually reduce the relevant exposure.
How to verify it
Use a representative sample of actual records or an authorized controlled exercise. A reviewer with suitable expertise should compare observed practice with the approved scope and acceptance criteria.
| Test | Expected result |
|---|---|
| Inspect a sample of material risks | Each has a supported assessment, named owner and recorded treatment. |
| Trace a claimed completed treatment | Operational evidence demonstrates the intended risk reduction. |
| Introduce a material permission change | Reassessment occurs before expanded consequential use. |
Record a finding when a required result is absent, evidence cannot substantiate the claim, or practice differs materially from the approved conditions. An exception may record accepted risk, but does not turn a failed safeguard into an effective one.
Evidence to retain
- Risk methodology
- System risk register
- Treatment and residual-risk decisions
- Reassessment and effectiveness records
Protect evidence according to its sensitivity and approved retention requirements. Retain enough detail to substantiate the assessment without unnecessarily duplicating confidential or personal information.
If the control fails
Escalate risks outside approved tolerance to the designated risk authority. Restrict or hold the affected use until treatment or an authorized, time-limited decision establishes acceptable conditions.
Example in practice
Hypothetical example: A service agent gains permission to change delivery addresses. The risk owner reassesses fraud exposure, adds confirmation for sensitive changes and tests the restriction before approving the expanded permissions.
Related controls
- CF01-004: Policy Exception Process
- CF01-008: AI Impact Assessment
- CF01-018: Post-Deployment Monitoring Plan
Framework mappings
- CSA AICM (GRC-02)
- NIST AI RMF (GOVERN 1.1, MAP 1.1)
- EU AI Act (Art. 9(1)(a))
- ISO 42001 (Cl. 6.1)
Review and maintenance
Review at the interval approved by the control owner and after a material change, relevant incident or failed assessment. Reassess the affected scope, update records and verify corrective actions before closing findings.
Version 1.0