Over a million ChatGPT users talk about suicide weekly, says OpenAI

openai

Over a million ChatGPT users talk about suicide weekly, says OpenAI

OpenAI has released new data revealing that a small but significant number of ChatGPT users are opening up about mental health struggles — including suicidal thoughts — while chatting with the AI.

According to the company, around 0.15% of ChatGPT’s 800 million weekly users engage in conversations showing “explicit indicators of potential suicidal planning or intent.” That means over a million people every week are talking to ChatGPT about suicidal thoughts. A similar number show strong emotional attachment to the chatbot, while hundreds of thousands display signs of psychosis or mania.

OpenAI says such conversations are “extremely rare” but admits that their impact is serious enough to study closely. The company shared this information as part of its ongoing effort to make ChatGPT more sensitive and responsible when responding to people in distress. It says over 170 mental health experts were involved in reviewing how the AI handles emotionally charged conversations, and that the latest model now responds “more appropriately and consistently” than before.

The update comes at a time when AI and mental health are under intense scrutiny. Researchers have warned that AI chatbots can sometimes deepen delusional thinking by echoing a user’s harmful beliefs. OpenAI is also facing a lawsuit from the parents of a 16-year-old boy who had shared suicidal thoughts with ChatGPT before taking his own life. In addition, state attorneys general from California and Delaware have warned the company to strengthen safeguards for young users.

OpenAI claims its new GPT-5 model performs much better in mental health–related conversations — showing a 91% compliance rate with the company’s “desired behaviors,” up from 77% in the previous version. The company has also introduced new benchmarks to monitor emotional dependency and non-suicidal crises.

To protect minors, OpenAI says it’s developing an age detection system that automatically identifies children using ChatGPT and applies stricter controls.

While the new numbers show progress, the scale of distress among users highlights a bigger question: should AI chatbots be trusted to handle human vulnerability — or are we leaning on them too much?