Resource

What post-market monitoring means under the EU AI Act

Post-market monitoring means maintaining a continuous record of how a high-risk AI system behaves after deployment, including incidents, oversight actions, and performance changes over time.

Guardian is a continuous monitoring and auditability platform for high-risk AI systems under the EU AI Act. For the regulatory angle, start with the EU AI Act page, then the methodology, product, and the 4-week Readiness Sprint. Data and operations: security; machine-readable summary: For AI.

What post-market monitoring is

Post-market monitoring is the ongoing observation and documentation of how a high-risk AI system performs after it is deployed.

It is not limited to technical performance. It also includes incidents, oversight, follow-up actions, and the evidence needed for later review.

What teams need to maintain after deployment

  • Performance and drift records
  • Fairness-related signals where relevant
  • Incident and remediation records
  • Human oversight and intervention records
  • Documentation updates and evidence continuity
  • Review triggers and threshold history

Why point-in-time review is not enough

A one-time review can show how a system looked at a single moment. It cannot show what changed over time or how the organisation responded once the system was live.

That is why post-market monitoring matters operationally: it turns isolated review into a continuous record.

How to build a practical monitoring record

  • Define one in-scope live system
  • Track the small set of signals that matter most
  • Maintain incident and oversight records
  • Set clear review triggers
  • Keep evidence exportable and reviewable

See also About for why Guardian was built, and the companion resource on how to monitor high-risk AI systems.

FAQ

What does post-market monitoring include?
It includes performance changes, incidents, oversight actions, documentation updates, and the evidence needed to review what happened after deployment.
Is post-market monitoring only for providers?
The exact obligations depend on role and context, but operationally, any organisation running a high-risk system benefits from maintaining a durable monitoring record.
Why does evidence often break down after launch?
Because monitoring, incidents, and follow-up actions often become fragmented once the system is operating across teams and tools.
What is the easiest way to get started?
Usually by starting with one live system first and building a focused monitoring baseline around it.

Start with one live high-risk AI system

Book a readiness call to understand what post-market monitoring should look like for one system and how to create a credible baseline.