Hallucination

A hallucination is when an AI states something false with complete confidence, an invented statistic, a fake citation, a feature that doesn't exist, a customer detail it made up. The model isn't lying; it predicts plausible-sounding text, and sometimes the most plausible continuation simply isn't true. The danger is that hallucinations read exactly like correct answers.

For a founder this is the single biggest risk in putting AI in front of customers or in your decisions. An agent that confidently quotes the wrong price, cites a non-existent policy, or fabricates a number can cost you trust or money before you notice. It matters because you can't eliminate it, only manage it: ground the model in real documents (RAG), keep a human checking anything high-stakes, ask for sources you can verify, and never let an unchecked agent take an irreversible action on a hallucinated fact.