Technical
Hallucination
When an AI model generates output that sounds plausible but is factually incorrect or fabricated.
In AI, a hallucination occurs when a language model generates information that is confident and coherent but factually wrong, fabricated, or unsupported by its training data. Hallucinations are an inherent characteristic of current LLM architectures — the models are optimized to produce fluent text, not necessarily accurate text.
Hallucination risks in enterprise contexts include:
- AI research tools presenting fabricated statistics or citations
- Coding assistants generating plausible but buggy or insecure code
- Customer-facing AI producing incorrect product information
- AI-generated legal or compliance content with fabricated regulatory references
- Vendor research reports that mix accurate and fabricated details
Why it matters
If an employee uses AI-generated content in a client deliverable, regulatory filing, or internal report, a hallucination becomes the organization's mistake. AI policies should address how AI-generated content is verified before use.