What Is Hallucination in AI 2026
Hallucination is when an AI confidently states something false. The biggest reliability problem with LLMs in 2026. Why It Happens LLMs are trained to predict plausible next tokens, not factual ones. They don't have a "I don't know" instinct — they fill in. Common Types - Factual hallucination: "The capital of Australia is Sydney" wrong — it's Canberra - Source hallucination: cites a paper that do…