Glossary

AI Hallucination

When an AI model states something confidently that is actually wrong or made up, rather than saying it doesn’t know.

A hallucination is a wrong or invented answer that an AI model presents as if it were fact. It’s one of the biggest risks of AI in a business setting, especially for anything involving prices, policies, or legal claims. AI Abstraction reduces hallucinations by grounding agents in a business’s real documents through RAG and testing responses before launch, rather than relying on a model’s general knowledge alone.

See every term in the glossary