AI Control Library / CF-14: Agentic & Autonomous System Controls
Designation: Core | Control types represented: Operational and Technical
Control requirement
Define task completion, cancellation, timeout and failure conditions, and implement graceful termination that reaches workers, queued actions, sessions and delegated tasks. Verify the final state rather than relying on a stop acknowledgement.
Why this matters
An interface may show cancellation while an external request is already executing. Safe termination requires distinguishing work stopped, work completed and effects still uncertain.
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 long-running, asynchronous, recurring and delegated workflows, especially those with external side effects.
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 workflow owner defines completion criteria; orchestration teams implement cancellation; service owners reconcile external effects.
Implementation guidance
- Define terminal states and deadlines for every workflow branch.
- Propagate cancellation and revoke task-scoped authority.
- Handle in-flight requests with idempotency, reconciliation and safe checkpoints.
- Test termination under network failure and worker unavailability.
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 |
|---|---|
| Cancel with queued actions | Unstarted actions cannot execute. |
| Worker loses connection | Task authority expires within defined limits. |
| External result uncertain | Reconciliation identifies outcome before retry. |
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
- State model
- Cancellation traces
- Timeout configuration
- Reconciliation 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
Prevent new work, isolate unreachable workers and reconcile uncertain effects before declaring the task stopped.
Example in practice
Hypothetical example: A travel agent is cancelled while booking a hotel. The workflow checks the provider’s transaction record before retrying or claiming that no reservation exists.
Related controls
- CF14-002: Agentic Action Reversibility Requirement
- CF14-007: Autonomous Replication Prevention
- CF14-009: Multi-Agent Orchestration Security
Framework mappings
- NIST AI RMF (GOVERN 5.1)
- MITRE ATLAS
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 Agentic & Autonomous System Controls
Version 1.0