Market surveillance authorities should be in a position to take all appropriate corrective measures where necessary to address any risks to health, safety, fundamental rights, the environment or public security caused by non-compliant AI systems.