Two new threat modes can flip generative AI model behavior from serving your GenAI applications to attacking them, according to three security researchers. While a jailbroken GenAI model itself may ...