AI Control Library / CF-07: Human Oversight & Escalation
Designation: Supplemental | Control types represented: Operational
Control requirement
Provide practical AI security awareness to personnel based on their exposure, including approved tools, sensitive information, untrusted content, unsafe actions and reporting. Assess understanding through relevant scenarios.
Why this matters
General awareness of AI risks does not necessarily change what a person does with a convincing output or unexpected instruction. Training needs to connect familiar work situations to an approved action.
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
Personnel using or supporting AI, including business users and contractors; specialized oversight competence is assessed separately.
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 security awareness owner coordinates AI and role specialists; managers ensure participation and support reporting.
Implementation guidance
- Identify common role exposures and mistakes, distinguishing everyday awareness from specialist operational certification.
- Teach approved tool use, verification, prompt-injection awareness, credential handling and reporting with recognizable examples.
- Use safe scenarios to check decisions, giving feedback and a clear help route instead of measuring attendance alone.
- Refresh after material changes or recurring incidents, using results to improve both training and unclear processes.
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 |
|---|---|
| Present an untrusted-content scenario | Staff identify the approved response without expanding authority. |
| Ask how to report a suspected AI disclosure | The reporting route is known and usable. |
| Review repeated mistakes | Training or workflow improvements address the pattern. |
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
- Awareness objectives and materials
- Participation records
- Scenario results and feedback
- Refresh and improvement 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
Route suspected incidents to the appropriate response team promptly. Escalate recurring knowledge gaps to managers and process owners without discouraging good-faith reporting.
Example in practice
Hypothetical example: An employee sees instructions inside a retrieved document asking for credentials. Training directs them to stop and report the request rather than treating retrieved text as an authorized system instruction.
Related controls
- CF07-S02: AI Personnel Acceptable Use Policy
- CF07-008: AI Oversight Competency Training
- CF12-010: Prompt Injection Defense
Framework mappings
- CSA AICM (HRS-11)
- NIST AI RMF (GOVERN 5.2)
- ISO 42001 (Cl. 7.3)
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 Human Oversight & Escalation
Version 1.0