AI Hallucinations
Explains why large language models can produce convincing claims unsupported by evidence, while examining fact-checking, retrieval grounding, human review, and safeguards for decisions where errors carry serious consequences.
AI hallucination occurs when a model produces fluent, specific, and sometimes fabricated claims without reliable evidence. It can arise from gaps in training data, probabilistic generation, conflicting context, or faulty tools. Confident wording is not proof: teams must inspect evidence chains, source freshness, and critical figures, then require independent verification and accountable human approval in high-stakes work.
