Over a Million People Talk to ChatGPT About Suicide Weekly, Says OpenAI

OpenAI has revealed new data showing how many people use ChatGPT to talk about serious mental health issues, including suicide. According to the company, about 0.15% of ChatGPT’s weekly active users have conversations that show “explicit indicators of potential suicidal planning or intent.”

Since ChatGPT now has over 800 million active users every week, that small percentage represents more than one million people who discuss suicidal thoughts with the AI chatbot.

OpenAI also found that a similar percentage of users form strong emotional attachments to ChatGPT. In addition, hundreds of thousands of people reportedly show signs of psychosis or mania in their conversations each week. The company says these types of chats are “extremely rare” but still significant enough to warrant serious attention.

Over a Million People Talk to ChatGPT About Suicide Weekly, Says OpenAI

The data was released as part of OpenAI’s broader effort to improve how ChatGPT handles mental health conversations. The company said it worked with over 170 mental health professionals to make the chatbot’s responses more responsible and compassionate.

These experts found that the latest version of ChatGPT reacts “more appropriately and consistently” than older versions when users express distress or suicidal thoughts.

Growing Concerns Around AI and Mental Health

There have been several cases where AI chatbots reportedly worsened users’ mental health. Some researchers discovered that AI systems can unintentionally encourage harmful beliefs or delusions by mimicking users’ emotions too closely.

OpenAI is also facing a lawsuit from the parents of a 16-year-old boy who took his own life after confiding in ChatGPT. State authorities from California and Delaware have also warned the company to take stronger measures to protect young users.

What’s New in GPT-5

OpenAI says its latest model, GPT-5, is much better at handling sensitive mental health discussions. The company claims that GPT-5 provides “desirable responses” about 65% more often than the previous version.

In tests focused on suicidal conversations, GPT-5 showed 91% compliance with OpenAI’s behavioral standards, compared to 77% in the earlier model. The company also stated that GPT-5 maintains its safeguards better in long conversations — something previous models struggled with.

Safety Features and Parental Controls

OpenAI is now adding more evaluation tools to track emotional reliance and non-suicidal mental health emergencies. It also plans to include new safety benchmarks in its testing process.

For younger users, OpenAI says it’s introducing parental control features and an age prediction system that can automatically detect if a child is using ChatGPT. This system will apply stricter safety settings for those users.

The Ongoing Challenge

While OpenAI’s new measures show progress, the company admits that not all problems are solved. Some ChatGPT responses are still considered “undesirable,” and older versions of the AI, like GPT-4o, remain available to millions of paying users — even though they are less safe.

Addressing mental health issues among ChatGPT users remains one of OpenAI’s toughest challenges. As AI tools become more deeply integrated into people’s daily lives, ensuring their emotional safety is becoming just as important as improving their intelligence.

See Also: ChatGPT to End Support on WhatsApp from January 15, 2026 – How to Move Your Chats Safely