- This topic has 0 replies, 1 voice, and was last updated 1 month, 1 week ago by
.
-
-
OpenAI revealed new data showing that more than a million ChatGPT users engage in conversations about suicide weekly. According to the company, roughly 0.15% of its 800 million weekly active users have interactions containing “explicit indicators of potential suicidal planning or intent.”
The data also shows that hundreds of thousands of users display signs of mania, psychosis, or heightened emotional attachment to the AI. OpenAI described these interactions as “extremely rare” but acknowledged that they affect a significant number of users each week.
This announcement comes alongside OpenAI’s efforts to improve how ChatGPT responds to mental health issues. The company says it consulted with over 170 mental health professionals to make the chatbot’s responses more consistent and appropriate.
OpenAI reports that its latest model, GPT-5, now produces “desirable responses” to mental health concerns about 65% more often than its predecessor. In testing suicidal conversations, GPT-5 met 91% compliance with OpenAI’s safety benchmarks, up from 77% for the previous version.
OpenAI’s focus on safety comes amid high-profile incidents. Earlier this year, the company faced a lawsuit after a 16-year-old reportedly confided suicidal thoughts to ChatGPT prior to his death. State attorneys general in California and Delaware have also warned OpenAI to implement stronger protections for minors.
In addition to model updates, OpenAI is rolling out more parental controls and an age prediction system designed to identify underage users and apply stricter safeguards automatically. The company is also introducing new safety benchmarks to measure AI performance in conversations involving emotional dependence or other mental health emergencies.
Despite these improvements, challenges remain. OpenAI continues to offer older, less-restricted models, including GPT-4o, to subscribers, meaning safety gaps could persist.
The numbers highlight a growing trend: for millions of people, AI chatbots are becoming confidants during moments of distress. OpenAI’s work on safety and mental health reflects an ongoing effort to balance accessibility with the need to protect vulnerable users.
-
- You must be logged in to reply to this topic.