OpenAI announced a partnership with the American Psychological Association (APA) on August 6, 2026, to bring psychological and developmental science into the safety work around young people’s use of AI. The premise is not that AI should replace mental-health care. It is that young people already use AI to learn, create, ask questions, and seek advice, while families, schools, and professionals need better evidence and practical resources.
OpenAI says the work will examine several areas: how AI can provide information responsibly in moments of distress, how experiences can be made more developmentally appropriate, and how parents, caregivers, clinicians, and educators can guide young people’s use. One planned output is family-facing guidance to help parents support healthy use, recognize when adult intervention may be appropriate, and keep AI in the role of a tool rather than a substitute for people or care.
Another part of the partnership focuses on practitioners and the broader field. OpenAI says it will support clinicians and school psychologists who are increasingly asked about AI, including how to encourage healthy use, recognize overreliance, and respond to unhealthy patterns. The company also plans to convene teens, families, clinicians, school psychologists, and others with lived experience so that safeguards are not designed only from product-team assumptions.
OpenAI connects the announcement to existing safety work. It says more than 260 mental-health experts help ChatGPT recognize signs of distress, respond with care, and guide people toward real-world support. The company also lists localized crisis resources, one-click hotlines, break reminders, parental controls, parent notification for safety concerns, under-18 principles, and an age-prediction model.
Those features and expert figures are OpenAI’s published product and safety descriptions, not independent evidence of clinical effectiveness. Youth AI safety is difficult for reasons beyond the model’s tone: age estimation can be wrong, crisis resources need regional fit, parent notifications can create privacy and safety tradeoffs, and systems need to know when to return a user to real human support.
The partnership is best understood as an evidence loop rather than a new chatbot feature. If research, families, schools, clinicians, and young people can genuinely influence product rules, AI safety has a better chance of becoming testable practice instead of abstract policy. The central boundary is clear: AI may help people find information and navigate support, but it should not replace family, professional care, or human relationships.
