Human oversight
What it means
What does human oversight require under the EU AI Act?
Human oversight is the EU AI Act requirement that a person watching a high-risk AI system can understand what it is doing, catch a wrong output and stop the system. That takes training, time and authority, not just a name on a workflow diagram.
Article 14 splits human oversight into two kinds of measure. The provider builds one kind into the system itself before it ships, such as an interface that flags a low-confidence output. The provider identifies the other kind for the deployer to carry out instead. An example of that kind is a rule about when a person must review a case before it proceeds.
Article 14 lists five capacities the overseeing person needs. They must understand the system's real limits well enough to spot a malfunction. They must stay alert to automation bias, the tendency to trust the output by default. They also need to interpret the output correctly and to disregard or reverse a specific output when needed. Last, they need a way to halt the system outright.
Published · Updated
In practice
Why it comes up
It differs from asking whether a human is in the loop. That is a design question about where a person sits in a workflow. Human oversight is a legal test of whether that person can catch a bad output and has the time, context and authority to stop it. An auditor tests this by asking what the overseer would do in a specific bad case, not whether a review step exists on a diagram.
What to do about it
Name the specific person or role authorised to halt each high-risk system, not a team. Check that the person has time per case to evaluate output rather than approve on reflex. Train that person specifically on automation bias. A rushed or over-trusting reviewer defeats the requirement even when the process looks correct. The Fractional CAIO Retainer includes reviewing who watches each system, and whether they could catch a failure this week.
Sources and neighbours
Sources
- Article 14 - Human oversight
eur-lex.europa.eu
- NIST AI Risk Management Framework
nist.gov