Hallucination
What does it mean?
A hallucination is an AI model's answer that sounds plausible but is factually incorrect or made up. Because a language model predicts language patterns rather than looking up facts, it can confidently state something untrue. Source citations, RAG and human review significantly reduce this risk.
Hallucinations are not a 'bug' you fully program away, but a property of how language models work. You manage them: have the model answer based on your sources (RAG), show the sources, and make sure employees know where human review is needed.
AI literacy plays a key role here: those who understand how a model reaches an answer recognise when a check is needed. That is exactly what our training instils.
From concept to application?
Book a no-obligation call. We translate these terms into what they concretely mean for your organisation.