AI Control Library / CF-01: Lifecycle Governance
Designation: Supplemental | Control types represented: Operational and Technical
A successful backup job does not prove the system can be restored. Recovery may depend on model artifacts, retrieval content, configuration, permissions and compatible software rather than application files alone.
Control requirement
Back up the artifacts and state needed to restore AI-supported services to an approved condition, with protection, retention and recovery testing proportionate to business needs and data obligations.
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 define the process and decision authority; Technical controls enforce boundaries or produce the system evidence needed to verify it.
When it applies
AI artifacts and deployment state under organizational control; document externally hosted components that cannot be backed up directly.
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 backup service owner operates protection; AI system owners identify required assets and verify service-level recovery.
Implementation guidance
- Derive backup scope and recovery-point requirements from the baseline and business impact assessment; include indexes, prompts, configuration and relevant state.
- Protect backup confidentiality and integrity, separate recovery access from ordinary administration where appropriate and retain usable recovery-key access.
- Monitor backup completion and retention, including deletion obligations and the risk of restoring data that should no longer be used.
- Restore representative backups in isolation and verify dependencies, data consistency, permissions and application behavior against the approved baseline.
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 |
|---|---|
| Restore a selected backup | Required service artifacts and compatible state are usable. |
| Assume a privileged account is compromised | Recovery copies and access remain protected under the defined threat model. |
| Reconcile retention and deletion | Recovery does not silently reintroduce prohibited data into service. |
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
- Backup scope and schedule
- Protection and key-access configuration
- Job and retention records
- Restore and data-reconciliation tests
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 missing or failed backups and unverified restore capability to system and continuity owners. Restrict risky changes or reliance where recoverability falls below approved needs.
Example in practice
Hypothetical example: A RAG service restores its application but lacks the retrieval index snapshot and document permissions. The test fails, prompting expansion of backup scope and a full service restoration test.
Related controls
- CF01-012: AI System Baseline Management
- CF01-013: AI System Rollback & Restoration
- CF01-S12: AI Disaster Response Plan
Framework mappings
- CSA AICM (BCR-08)
- NIST AI RMF (MANAGE 4.1)
- ISO 42001 (Cl. 8.2)
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