Article 72: Post-Market Monitoring — Keeping Watch After Deployment

What Is Post-Market Monitoring

Post-market monitoring means keeping watch over an AI system after it's been deployed — not just testing it once and forgetting about it. Just like pharmaceutical companies monitor drugs after they hit the market, AI providers must track how their systems perform in real-world use. This ongoing surveillance helps catch problems that weren't apparent during development and testing.

The requirement recognizes that AI systems can behave differently in the real world than in test environments. Data changes, user behavior varies, and unexpected situations arise. Without monitoring, problems can grow before anyone notices.

Provider vs. Deployer Responsibilities

The primary obligation for post-market monitoring falls on the AI provider — the company that built and sells the system. They must create a monitoring plan, collect performance data, and analyze it for signs of problems. However, deployers (businesses that use the AI) play a crucial role in this system.

As a deployer, you're the one seeing the AI in action every day. You're often the first to notice when something doesn't seem right. The EU AI Act expects you to cooperate with the provider's monitoring efforts by reporting issues, sharing relevant data, and following up on provider recommendations.

What to Watch For

In your day-to-day use of AI tools, pay attention to patterns of incorrect outputs, unexpected or nonsensical results, situations where the AI seems to perform poorly for certain groups of people, security incidents or data breaches involving the AI system, and user complaints about AI-assisted decisions.

When you spot something concerning, document it and report it to your AI vendor. Keep your own records of issues you've encountered — this is both good business practice and a compliance requirement.

Building Monitoring Into Your Operations

Make AI monitoring part of your regular business routines. Set up a simple process for staff to flag AI issues. Review AI performance in team meetings. Track key metrics like accuracy rates and error types. This doesn't need to be complex — a shared spreadsheet where team members log AI problems is a good start. The important thing is that you're systematically watching, not just hoping everything is fine.

This article is for informational purposes only and does not constitute legal advice. Regulatory requirements change frequently — verify current rules with official sources. Built by Sawai Gyoseishoshi Office, Hiroshima, Japan.