Human-in-the-loop refers to processes where human oversight is required to review, validate, or approve AI outputs before decisions are made or actions are taken.