Hallucination

When a model states confident but false info because it predicts plausible language, not truth.