AI Control Library / CF-12: Security & Abuse Resistance
Designation: Core | Control types represented: Operational and Technical
Control requirement
Authorize and trace material modifications to AI-generated outputs before they are approved, published or used in consequential processes. Distinguish original generation, edited versions and final authorization.
Why this matters
An accurate model output can become misleading after manual or automated editing. Downstream users need confidence that the version acted upon is the version approved.
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 output editing, post-processing, publication and transfer into business records. Scale approval to the consequence of the change.
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 output process owner sets modification authority; application teams enforce editing rights; reviewers approve consequential versions.
Implementation guidance
- Define who may edit, approve and publish each output category.
- Preserve version relationships and meaningful change history.
- Bind approval to the exact final content and target.
- Reopen review when material post-approval changes occur.
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 |
|---|---|
| Unauthorized edit | Application denies modification. |
| Approved output changes | Prior approval does not silently carry forward. |
| Published sample | Final version and authorization are traceable. |
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
- Editing role matrix
- Version history
- Approval records
- Publication traces
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
Hold or withdraw the altered output, identify downstream use and obtain valid review before republishing.
Example in practice
Hypothetical example: An analyst edits an AI-generated safety notice after approval. The publishing system requires review of the changed wording before the notice can be released.
Related controls
- CF12-S03: AI System Separation of Duties
- CF12-S12: AI System Authorization Mechanisms
- CF04-001: AI Audit & Assurance Policy
Framework mappings
- CSA AICM (IAM-18)
- 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