AI Control Library / CF-07: Human Oversight & Escalation
Designation: Supplemental | Control types represented: Operational
Control requirement
Establish and administer appropriate confidentiality agreements for AI personnel with access to protected information, clarifying covered material, permitted handling, reporting and applicable limits through qualified review.
Why this matters
AI work can expose prompts, model artifacts, customer records and supplier material. Confidentiality terms should address those realities while avoiding the implication that an agreement replaces access controls or prohibits lawful disclosures.
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
Employees, contractors and other authorized personnel with relevant organizational or third-party confidential AI information.
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
Legal and HR approve terms; role and information owners ensure coverage and handling instructions.
Implementation guidance
- Identify protected information and relevant personnel relationships, checking existing agreements before adding new terms.
- Define covered material, permitted use, authorized recipients and return or disposal responsibilities in reviewed language.
- Address applicable exceptions, reporting routes and continuing obligations through qualified legal review.
- Record acceptance and changes, and coordinate access, training and departure handling with agreement scope.
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 |
|---|---|
| Trace a sensitive-role assignment | Appropriate reviewed confidentiality coverage exists. |
| Present an incident-reporting scenario | Personnel know the authorized route and do not treat confidentiality as a reason to conceal the issue. |
| Inspect departure handling | Relevant information obligations and access actions are coordinated. |
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
- Confidentiality scope and approved terms
- Acceptance and version records
- Handling and reporting instructions
- Departure and coverage reviews
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
Escalate suspected disclosure to security and the information owner through the incident process. Refer term interpretation and personnel decisions to legal and HR.
Example in practice
Hypothetical example: A model evaluator receives confidential customer prompts and supplier artifacts. Reviewed terms define permitted handling, while protected workspaces and access controls enforce the operational boundary.
Related controls
- CF07-S07: AI Personnel Agreement Process
- CF07-S08: AI Personnel Agreement Content
- CF05-012: AI Sensitive Data Protection
Framework mappings
- CSA AICM (HRS-10)
- NIST AI RMF (GOVERN 5.2)
- ISO 42001 (Cl. 7.2)
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 Human Oversight & Escalation
Version 1.0