Human oversight (human-in-the-loop)

Human oversight covers several designs: human-in-the-loop (a person approves each decision), human-on-the-loop (a person supervises and can intervene) and human-in-command (people decide whether and how the system is used at all). The EU AI Act (Art. 14) requires high-risk systems to be designed so that assigned people can understand capacities and limitations, remain aware of automation bias, correctly interpret outputs, decide not to use or to override them, and interrupt the system. Deployers must assign oversight to people with the necessary competence, training and authority.

Oversight must be real, not a rubber stamp. Under GDPR Art. 22, a decision is only outside the automated decision-making rules if a person with authority meaningfully reviews it. Effective oversight depends on explainability, AI literacy and workloads that leave time to question the system.

Example: in a radiology workflow, the AI flags suspicious scans, but a radiologist reviews every case and records disagreements, which feed back into monitoring.