hallucination
When an AI model states something false with full confidence.
Language models generate text that sounds right, based on patterns. They do not check facts against a source unless they are built to. So they sometimes invent a figure, a quote or a reference that looks real but is wrong. This is called a hallucination. It is why important outputs still need checking.
Think of it as
Like a confident dinner guest who fills a gap in their memory with a plausible story, and tells it as if it were fact.
Example
Asked for sources on a legal question, a chat assistant lists three court cases with names and dates. One of them does not exist. A lawyer who files it without checking has a real problem.