George Washington University Researchers Create Formula to Anticipate Harmful AI Chatbot Outputs

Researchers at George Washington University have developed a mathematical formula designed to predict when public AI chatbots might generate harmful responses, such as those related to self-harm, extremism, or financial misinformation. This research aims to provide AI model developers with a tool to better understand and prevent undesirable outputs, enhancing safety and reliability.