MIT Develops New AI Red-Teaming Method to Enhance Chatbot Safety

Researchers have created a machine learning technique that generates more diverse prompts to better test and secure AI chatbots against toxic responses.