AI Control Library / CF-09: Monitoring & Drift Detection
Designation: Core | Control types represented: Operational and Technical
Control requirement
Update AI threat detections using relevant new evidence, incidents, advisories and system changes. Validate coverage, false positives and operational impact before and after deploying revised detection logic.
Why this matters
A detection tuned for one tool interface may fail after an integration change. New threat information should lead to an applicability decision and tested update rather than an unreviewed rule copied into production.
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
AI security detection rules, analytics, signatures, baselines and response routing across the monitored system.
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 detection engineering owner coordinates threat analysis, system specialists and security operations.
Implementation guidance
- Collect credible threat and incident inputs, recording applicability to the organization’s models, tools and exposure.
- Translate relevant findings into detection cases with required events, expected signals and known limits.
- Test revised logic with safe adversarial and benign cases, controlling rollout and retaining versions.
- Review live effectiveness and operating impact, correcting excessive noise and measuring whether relevant paths remain uncovered.
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 relevant advisory | Applicability and detection action or reasoned exclusion are recorded. |
| Exercise an updated rule | Relevant signals are detected without unacceptable false positives. |
| Change a required event schema | The detection is revalidated rather than silently losing coverage. |
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
- Threat applicability decisions
- Versioned detection logic
- Adversarial and benign tests
- Rollout and effectiveness 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 newly uncovered material threats to security and system owners. Apply interim restrictions or manual review while validated detection and prevention are established.
Example in practice
Hypothetical example: A new plugin changes action event fields. Detection testing discovers that the old permission-abuse rule no longer matches and repairs coverage before the integration expands.
Related controls
- CF09-005: AI Security Monitoring & Alerting
- CF08-002: AI Threat Analysis & Modelling
- CF04-008: AI System Logging Scope & Event Definition
Framework mappings
- CSA AICM (TVM-03, TVM-04)
- NIST AI RMF (MANAGE 4.1)
- MITRE ATLAS
- OWASP LLM Top 10
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 Monitoring & Drift Detection
Version 1.0