There is a Chatbot AI product called character.ai that has suffered a marked decline in quality since its launch as they battle their users to maintain the AI’s safety protocols (similar to chatGPT “jailbreaks”). I wonder if something similar could be happening here.