noun
An AI hallucination is when a model produces incorrect, fabricated, or nonsensical content that it presents as factual. This happens because AI systems generate responses based on patterns in their training data, not on verified knowledge.
Hallucinations often occur when the AI lacks sufficient context, is asked to fill gaps in data, or is pushed beyond its training limits. For example, a language model might invent a fake scientific study with citations that don’t exist, or an image generator might create a person with impossible anatomy.
For users, this means AI outputs must always be fact-checked, especially in high-stakes fields like medicine or law. Developers must design systems to flag potential hallucinations, often using human oversight or additional verification tools. A 2023 study found that 15% of AI-generated legal briefs contained fabricated case references.
A chatbot claims a celebrity was born in 1980 when records show 1978.
In other languages
See also: AI bias · prompt engineering · Model training