hallucination할루시네이션
When an AI confidently makes up something that isn't true.
Because an LLM predicts plausible words, it can invent things it doesn't know — citing papers that don't exist or stating wrong numbers are classic examples.
Asking for sources and checking important points against the original is the most reliable defense. Supplying reference material, as RAG does, also helps.
For example
Quoting a legal clause number that doesn't actually exist.