The practice of including human review or intervention at a point in a generative AI system's workflow, especially before high-risk or high-impact actions, as an additional safeguard against model errors or hallucinations.
In GENAI-L (Domain 3, section 3.1) it is listed alongside grounding, RAG, prompt engineering, and fine-tuning as a Google Cloud best practice for mitigating foundation model limitations.
HITL (Human in the Loop) is a concept that can show up on the exam. Check the full guide to see it in context.