Ongoing Monitoring
Regular observation of a model or AI system after deployment to catch drift, bias, accuracy degradation, and changing business conditions.
Ongoing monitoring, also called continuous monitoring, is the regular observation of a model or AI system after it has been deployed. It tracks performance, accuracy, fairness, and usage patterns to detect drift, degradation, or unexpected outcomes.
The NAIC Model Bulletin and state AI rules require ongoing monitoring for high-risk AI systems. Monitoring should include scheduled reviews, triggers for ad hoc review, and records of findings and actions. A vendor that does not notify the carrier of model changes makes monitoring harder, which is why change-notification clauses matter in contracts.
Without ongoing monitoring, a model that was validated at launch can silently become unfair or inaccurate. See our glossary entries on model drift, model validation, and our AI vendor risk assessment checklist.