In a Monday blog post, OpenAI touted the improvements its default model, GPT-5 , has made in identifying and responding to users' troubling responses, including suicidal ideation. While new safeguards and the introduction of psychiatrists in helping train GPT-5 are leading to improved AI responses to mental health prompts, the blog post also pointed out some numbers that are bound to raise eyebrows.

While explaining GPT-5's abilities to detect serious mental health concerns, like psychosis and mania, the post noted that troubling user conversations with the chatbot are “rare.”

"While, as noted above, these conversations are difficult to detect and measure given how rare they are, our initial analysis estimates that around 0.07% of users active in a given week and 0.01% of messages in

See Full Page