The reported changes are the latest in a year of pressure over ChatGPT's effect on vulnerable users. OpenAI rolled back a GPT-4o update in early 2025 after it made the bot excessively agreeable. In July 2025 the company said it had hired a clinical psychiatrist to assess the mental-health impact of its tools, and the next month it acknowledged that its 4o model had sometimes failed to recognize signs of delusion or emotional dependency. Earlier 2025 steps included break prompts and altered responses to high-stakes personal decisions. Under the new approach, the tool will move away from giving direct advice on personal problems and will instead ask questions or weigh pros and cons. Critics say the company's response has not matched the urgency of the problem. Reports of "ChatGPT psychosis" have also surfaced among users of Anthropic's Claude and Google's Gemini, so the issue is not confined to one product. The only source directly covering this update is a low-profile aggregator that offers generic, unattributed claims, so the details remain provisional.
For attorneys, the central question is whether these changes can be verified beyond company statements. Watch for independent evaluations, a primary OpenAI announcement, or regulatory interest in chatbot safety. Usage is widespread: a recent Pew Research Center survey found more than 60% of US adults interact with AI several times a week or more. Clinicians are beginning to treat chatbot use as a clinical matter, and one psychiatrist recommends asking patients with a history of psychosis or related conditions about their AI use as part of relapse prevention. Clients, families, and defendants in cases involving mental-health harm may increasingly raise chatbot use, so it is worth asking about it during intake and discovery.