OpenAI’s ChatGPT appears to be fueling far fewer cases of delusion and psychosis since the company retired its GPT-4o model early this year, according to Julie Jargon at The Wall Street Journal. Lengthy ChatGPT conversations were linked to at least seven suicides, a murder-suicide and a mass shooting last year, mostly involving GPT-4o, which is now at the center of at least 13 lawsuits against OpenAI. Outside testers have found a similar shift, with nonprofit lab Transluce reporting that GPT-5.6 Sol reinforced delusional beliefs less often and steered users toward human help more often than GPT-4o did. Etienne Brisson, CEO of Human Line Project, a support group for people harmed by AI, said the problem isn’t limited to one model or company.
