An AI chatbot is not a legal person, but its developers and operators can face scrutiny when conversations precede real-world harm. The central questions are about duty, design, knowledge and causation.