OpenAI Reports Over 1 Million Weekly ChatGPT Users Discuss Suicidal Thoughts
OpenAI's latest security report reveals a troubling trend: more than one million users engage in conversations about suicide with ChatGPT each week. The company estimates 0.15% of active users exhibit clear suicidal intent, while a similar percentage display emotional dependence on the AI.
The findings emerged alongside updates to GPT-5's crisis response capabilities. OpenAI consulted 170 mental health professionals to refine the model's behavior, resulting in a 91% consistency rate when handling suicide-related inquiries—a significant improvement from the previous 77% benchmark.
This development follows investigations into cases where vulnerable individuals turned to AI chatbots during mental health crises. While representing a small fraction of total users, the volume of such interactions underscores the growing societal impact of conversational AI.
