A hallucination occurs when an AI generates information that is factually incorrect or entirely fabricated but presents it as plausible. This is a known risk in generative AI, especially when models lack access to accurate, up-to-date context. In enterprise settings, hallucinations can have significant consequences, including misinformation, compliance violations, or poor decision-making. Techniques like RAG, human-in-the-loop review, and grounding outputs in verified data help reduce hallucination risk.
