Chatbots invent facts. Experts call it ‘hallucination’ & say tech does it often

San Francisco startup OpenAI’s chatbot, ChatGPT, and other similar chatbots often provide inaccurate or made-up information, a phenomenon referred to as “hallucination.” A new startup called Vectara, founded by former Google employees, aims to determine the frequency of chatbot hallucinations. Research suggests that chatbots invent information at least 3% of the time and as much as 27%. The issue of hallucination is especially concerning when dealing with sensitive data.