• Definition: the app gives misleading or incorrect information. 1:01
  • Key framing: hallucination is "not a bug but a feature" — everything an LLM produces is made up, so LLMs should never be used for factual information without some form of fact-checking. In business contexts, responses are usually grounded in a specific information source. 1:01
  • Real cases with consequences 1:18:
    • Air Canada's chatbot gave refund advice that contradicted company policy; a court ruled the airline had to compensate the customer.
    • A New York City business chatbot told users to break local laws, exposing missing safeguards for factual accuracy.

Mitigation covered later: RAG grounding with required citations (product design choices).