OpenAI Strengthens Sensitive Conversation Protection and Youth Safety

OpenAI announced new safety measures and feature improvement plans to make ChatGPT a more helpful experience -- to be intensively pursued over the next 120 days including crisis response expansion, expert collaboration strengthening, and youth protection feature enhancement. Two notable elements: automatically routing sensitive conversations to "reasoning models"; and launching comprehensive Parental Controls within one month. Expert advisory structure: (1) Well-Being and AI Expert Panel -- youth development, mental health, and human-computer interaction experts providing clear, evidence-based vision for how AI can support wellbeing; research findings to be directly reflected in future parental control features and safety mechanism design; (2) Global Physician Network -- 250+ physicians active in 60 countries, 90+ of whom are psychiatry, pediatrics, and family medicine specialists; their feedback directly applied to model training and safety research. AI routing for sensitive conversations: mental health, self-harm, suicidal ideation, and other high-risk conversations to be automatically identified and routed to o3/o4 reasoning models that have been specifically trained with additional safety fine-tuning. Parental controls architecture: age verification for ChatGPT accounts; usage limits and content filtering settings parents can configure; transparency reports to parents about account activity; the design challenge is providing useful controls without creating surveillance infrastructure that damages parent-child trust.