AIsince 2022⏱ ~1 min

Hallucination

When an AI model produces seemingly credible answers that are factually wrong, fabricated or without basis in reality.

Hallucination occurs because language models are trained to generate the most likely next text - not to verify facts. The model can "make up" references, names, legal paragraphs or technical specifications that sound plausible but do not exist. Hallucinations are one of the most important risks when using AIArtificial Intelligence - technology that enables machines to perform tasks that normally require human intelligence, such as language understanding, pattern recognition, and decision-making. in critical contexts. Techniques such as RAGRetrieval Augmented Generation - AI architecture where the model is supplemented with documents retrieved in real-time to provide fact-based answers. and tool use reduce the risk by grounding answers in verifiable data sources.

AIRisikoDatakvalitet