New Formula Predicts AI Chatbot ‘Bad Turn’ Behavior

Researchers developed a formula to anticipate when AI chatbots might exhibit undesirable or harmful behaviors, enabling proactive intervention and safety measures.