What is Human-in-the-Loop?
Human-in-the-Loop (HITL) is a design pattern where an AI system performs a task, but the result is verified and/or approved by a human before further processing. The human isn't replaced — they serve as quality and safety controller.
Three oversight variants
Human-in-the-loop: AI proposes, human accepts/rejects each decision. Human-on-the-loop: AI operates autonomously, human monitors and intervenes when something goes wrong. Human-in-command: human makes strategic decisions, AI provides analyses and recommendations.
When is HITL required?
The AI Act requires human oversight for high-risk systems (Art. 14). Beyond legal requirements, HITL is essential when: decisions have serious consequences, the model operates in new domains, AI errors could cause reputational damage.