We use cookies and similar storage that the site needs to work. With your permission, we would also like to use marketing cookies to understand which of our pages and campaigns interest you and, if we know who you are, to let our team follow up. You can change your mind at any time from Cookie settings at the foot of every page. Cookie Policy
Article 72: Post-Market Monitoring Without a Second Compliance Programme
Standard Intelligence
Post-market monitoring is the obligation that quietly outlasts every other part of AI Act compliance. Article 72 requires providers of high-risk systems to keep watching once the system is live, and to do so through a documented plan. The instinct is to stand up a new programme for it. That instinct is wrong.
A monitoring plan that lives apart from the rest of the business is the one that decays. The systems you already run, telemetry, incident handling, model performance review, change control, generate most of what Article 72 wants. The work is to connect them into a plan, not to build a parallel structure that competes for the same attention and budget.
A monitoring plan that lives apart from the rest of the business is the one that decays.
What Makes a Plan Defensible
So what makes a plan defensible? It states what is monitored and why, sets the thresholds and signals that matter, and defines who acts when a signal fires. It ties monitoring to the risk management of Article 9 rather than treating it as a separate exercise. And it records decisions, so an authority can see not just that you monitored, but that you responded.
The review triggers are where many plans are thin. A defensible plan names the conditions that force a reassessment of the system's documentation, performance drift, new failure modes, changes to the intended purpose or the deployment context, and links each to an owner and a timeframe. Without those triggers, monitoring becomes passive data collection that proves nothing.
Common Deficiencies
The common deficiencies are predictable. Plans that describe tools rather than decisions. Plans with no escalation path. Plans that gather metrics no one reviews. Plans that never feed back into the technical file. Each of these reads, to a supervisor, as monitoring on paper only.
The supervisory landscape is tightening as national authorities move from designation to operation, and post-market evidence is exactly the kind of thing a market surveillance authority can ask for without warning. A plan that is already wired into live processes answers that request. A plan that exists only as a document does not.
Our whitepaper, Post-Market Monitoring Plans, sets out the content of a defensible plan, the integration points with your existing infrastructure, the trigger conditions worth naming, and the deficiencies to design out. Monitoring is not administration. It is part of how a high-risk system stays lawful.
Global AI regulation will not stay fragmented. This paper forecasts its coalescence around three regulatory blocs on staggered timelines, and the bipolar interregnum that precedes the full structure.
Ungoverned AI is now a top-tier institutional risk. This paper sizes the exposure across thirteen categories and shows where governance investment produces the largest reduction in expected loss.
A single AI system rarely sits under a single rulebook. This paper sets out TRACE, a five-phase methodology for governing AI across overlapping frameworks, and a maturity model for locating where an organisation stands.