Which is good to some extent. We have people off the deep end sharing “unlock/jailbreak prompts” which turn LLMs in to schizophrenia machines affirming any psychosis. While a real person would push back and try to get you help.

ChatGPT can’t tell the difference between being given a harmless instruction / role play prompt, vs someone who is going insane. Probably explains why many of the most vocal AI users seem detached from reality, it’s the first time they have someone who blindly affirms everything they think while telling them they are so smart and completely correct all the time.