AI Control Library / CF-07: Human Oversight & Escalation
Designation: Core | Control types represented: Operational
Control requirement
Define and operate escalation paths for AI uncertainty, unexpected behavior, policy conflicts and failed safeguards, with actionable triggers, accountable recipients, response authority and fallback when the normal route is unavailable.
Why this matters
A contact list is not an escalation process if operators do not know when to use it or whether anyone can act. Triggers must connect observed conditions to timely decisions and containment.
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
AI operation, user support, monitoring and review where problems require higher authority or specialist judgment.
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 system owner defines escalation with operations, security and domain owners; designated responders maintain coverage.
Implementation guidance
- Identify conditions requiring escalation, including uncertain consequential output, denied actions, missed oversight and control failure.
- Assign recipients, severity, response expectations and authority to restrict or stop the affected workflow.
- Provide protected context and a reachable alternate route, distinguishing urgent containment from later analysis.
- Exercise handoffs and unavailable responders, reviewing unresolved cases and recurring triggers for process improvement.
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 |
|---|---|
| Generate a representative escalation | The correct role receives useful context and acts. |
| Make the primary responder unavailable | An alternate route or safe restriction works. |
| Review an overdue case | Escalation and containment do not depend solely on an unanswered message. |
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
- Trigger and severity matrix
- Response and coverage arrangements
- Handoff tests
- Case and overdue-action 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
Use the alternate route and approved containment when the initial path fails. Notify the accountable owner and keep consequential activity restricted where the unresolved issue exceeds tolerance.
Example in practice
Hypothetical example: An agent detects conflicting customer instructions before a payment. The task is held for an authorized specialist, with an alternate on-call route if the usual reviewer cannot respond.
Related controls
- CF07-001: Human Oversight Policy & Procedures
- CF07-006: AI System Kill Switch & Emergency Stop
- CF10-005: AI Incident Triage Process
Framework mappings
- CSA AICM
- NIST AI RMF (GOVERN 5.1, MANAGE 3.1)
- EU AI Act (Art. 14(1))
- ISO 42001 (Cl. 8.2, A6.1)
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