AI Control Library / CF-12: Security & Abuse Resistance
Designation: Supplemental | Control types represented: Operational and Technical
Control requirement
Separate incompatible AI duties so that one person or identity cannot independently initiate, approve and conceal a consequential change. Use independent review and compensating controls where staffing limits separation.
Why this matters
Combining development, deployment, access approval and audit alteration can allow errors or misuse to escape detection. Role names matter less than effective permissions.
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
Apply to model release, sensitive data use, privilege approval, output publication and evidence administration. Assess emergency workflows as well as routine operations.
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 process owner defines incompatible duties; identity teams enforce roles; risk owners approve documented exceptions.
Implementation guidance
- Identify consequential decisions and conflicting permissions.
- Separate request, approval, execution and independent review where needed.
- Enforce conflicts across systems and service identities.
- Review exceptions, emergency access and retrospective evidence.
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 |
|---|---|
| Self-approval attempt | Independent approval is required. |
| Cross-system role combination | Conflict is detected and resolved. |
| Emergency exception | Use is time-bound and independently reviewed. |
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
- Duty matrix
- Role configuration
- Exception approvals
- Independent review 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
Remove conflicting privileges or add an approved compensating review, then examine changes made during the conflict.
Example in practice
Hypothetical example: A model developer can propose a production release but cannot approve it or modify the protected release audit trail.
Related controls
- CF12-S06: AI Privileged Access Role Segregation
- CF12-S09: AI Audit Log Infrastructure Protection
- CF12-005: AI Output Modification Authorization
Framework mappings
- CSA AICM (IAM-04)
- NIST AI RMF (MAP 1.1)
- ISO 42001 (Cl. 8.2)
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 Security & Abuse Resistance
Version 1.0