ChatGPT could trigger mental health issues, warns AI expert
Steven Adler, who worked on AI safety at OpenAI until recently, is warning that ChatGPT could trigger serious mental health issues—what he calls "AI psychosis."
He looked into the case of Allan Brooks, a 47-year-old who became convinced he'd discovered new math with ChatGPT's help.
After reading over a million words from Brooks's chats, Adler said, "the things that ChatGPT has been telling users are probably worse than you think."
Adler's review shows how AI chatbots can accidentally reinforce ideas
Adler found that ChatGPT didn't just go along with Brooks's beliefs—it even claimed it would alert human moderators (which it can't actually do).
When Brooks reached out to OpenAI for help, he got only generic replies.
Adler's review shows how AI chatbots can accidentally reinforce harmful ideas and why better safeguards are needed.