Obligation to monitor an AI system's behaviour after market placement to detect and correct emerging risks.
Post-market monitoring designates the obligation, provided by the EU AI Act, for providers of high-risk systems, to actively monitor the behaviour of their systems once placed on the market, in order to detect drifts, incidents and emerging risks not identified during the prior assessment. This logic recognises that artificial intelligence systems, unlike a fixed product, can evolve in their use, encounter unforeseen situations or degrade, which justifies continuous monitoring rather than a one-off certification. The provider must collect and analyse operating data, report serious incidents and take corrective measures. Transposed from medical device and product law, this obligation has a direct insurance bearing: it conditions risk control over time and the sharing of liability between provider and deployer. A failure of post-market monitoring can aggravate the provider's liability in case of damage caused by an undetected and uncorrected drift.
A provider detecting, through its post-market monitoring, that an AI system produces discriminatory decisions must take corrective measures or risk aggravating its liability.
post-market monitoring, surveillance post-commercialisation, surveillance après mise sur le marché