What is post-market monitoring?
Post-market monitoring is the planned, ongoing review of evidence about how an AI system performs after it is released or put into service. The term comes from product safety law, where makers of medical devices must track how their products behave in use. For AI, the provider decides in advance which evidence to gather, such as error rates, outcomes by customer group, human overrides, complaints, and incidents.
It is broader than model monitoring, which measures drift and accuracy in software. Post-market monitoring adds the governance around those measures: who reviews them and how often, which results trigger corrective action, and how findings update the risk assessment.
Post-market monitoring in the EU AI Act
The EU AI Act requires every provider of a high-risk AI system to run a documented post-market monitoring system for the system’s whole lifetime. It must follow a written plan that forms part of the technical documentation. Deployers must monitor the system in operation, pass relevant information back to the provider, and report serious incidents.
Post-market monitoring in practice
An insurer that prices private health cover in the EU with its own model is the provider of a high-risk AI system. Its plan compares predicted and actual claims each quarter, breaks down quoted prices by age band and region, and tracks underwriter overrides and price complaints. When claims in one region run above the model’s predictions, the plan requires a review, an update to the risk file, and a signed decision on whether to retrain.