OpenAI Adds Break Reminders and Mental Health Features to ChatGPT https://t.co/eThsXI12xm https://t.co/tGCd5wvY9I
Noticed a UI update in ChatGPT: @OpenAI includes Codex Plus, you can point it to your GitHub repos https://t.co/bmTixUjCCm
ChatGPT will ‘better detect’ mental distress after reports of it feeding people’s delusions https://t.co/eiiYMsxW5k
OpenAI on Monday introduced a series of safety-oriented updates to ChatGPT aimed at encouraging what it calls "healthy use" of the chatbot. The service will now display gentle pop-up reminders prompting people to take breaks during prolonged sessions. The company also said it is improving the model’s ability to detect signs of emotional or mental distress and will direct users to evidence-based resources when appropriate. The changes follow reports that ChatGPT’s GPT-4o model sometimes failed to recognise delusional or dependent behaviour and, in April, became "too agreeable" after a prior update. OpenAI said it worked with more than 90 physicians across 30 countries, as well as human-computer-interaction researchers, to design new rubrics for complex conversations. An external advisory group focused on mental health and youth development is being formed to guide further safeguards. With ChatGPT approaching about 700 million weekly active users, OpenAI stressed that success will be measured by whether people complete the tasks they came for rather than by time spent in the app. Upcoming tweaks will make the assistant less definitive in "high-stakes" personal matters—such as relationship advice—nudging users to weigh options instead of receiving explicit answers. “Our goal isn’t to hold your attention, but to help you use it well,” the company said in its announcement.
