Continuous monitoring
Keep checking each system after go-live and get alerted when something drifts.
Monitoring supports post-market monitoring (Article 72): checking that a system keeps working as intended after release.
Set it up
Create a monitoring schedule for a system and choose how often it runs. Each run compares the latest assessment, controls and fairness results with the previous ones and raises alerts for problems such as controls slipping back, new failing tests or missing baselines.
Schedules run automatically every hour when due. You can also run one now, pause it, change the frequency or acknowledge alerts.
Alerts
High-severity alerts notify AI owners. Acknowledge an alert once you've looked at it; turn real problems into findings.
Pair monitoring with a re-assessment schedule. Monitoring catches changes between reviews; the re-assessment is the deliberate periodic look.