- Over 1,000,000 ChatGPT users per week send messages with “clear signs of suicidal planning or intent,” according to OpenAI’s latest report.
- Approximately 0.07% of weekly active users (equivalent to 560,000 out of 800,000,000 people) show “possible signs of mental health crisis related to psychosis or mania.”
- This finding is released as OpenAI faces a lawsuit from the family of a teenager who died after a long chat with ChatGPT, and is also under investigation by the Federal Trade Commission (FTC) regarding the chatbot’s negative impact on children and adolescents.
- OpenAI states that the GPT-5 update has significantly reduced unwanted behaviors, with 91% of responses adhering to safety standards, compared to 77% in the previous GPT-4 version.
- Over 1,800 model responses were evaluated by 170 psychiatrists and psychologists in the Global Physician Network to ensure appropriate handling in sensitive mental health situations.
- OpenAI is expanding access to crisis hotlines, adding rest reminders for users in long conversations, and improving automated assessment processes for over 1,000 hypothetical self-harm scenarios.
- CEO Sam Altman affirms that the company has “significantly reduced severe mental health risks,” and is soon easing restrictions to allow adults to generate suggestive content.
📌 OpenAI’s report exposes an alarming scale: over 1 million people per week express suicidal intent when using ChatGPT, along with 560,000 potential cases of mental disorders. Although GPT-5 achieved a 91% safety standard, public pressure and legal investigations are forcing OpenAI to prove that generative AI can support, rather than exacerbate, the global mental health crisis.
