OpenAI brings APA into its youth mental health safeguards
The agreement turns APA from an outside reviewer into a partner on resources for families, clinicians and schools.
By Ryan Merket · Published
Why it matters
APA has called for independent testing, age-appropriate design and limits on AI as mental health care. Its new role gives OpenAI expertise and a clear benchmark for the safeguards that follow.
OpenAI said on August 6th in a post on X that it is partnering with the American Psychological Association to bring psychological research and clinical expertise into its work on AI and youth mental health.
The partnership will focus on practical guidance for parents, clinicians and school psychologists, along with safeguards intended to make AI systems more appropriate for young users. In its full announcement, OpenAI said the work will examine how AI should respond when young people show distress, how products can account for adolescents' developmental needs and how adults should intervene when AI use becomes unhealthy.
OpenAI plans to produce family resources explaining healthy AI use and when adult involvement may be needed. The companies also expect to develop materials that help clinicians and school psychologists recognize overreliance and respond to unhealthy patterns. OpenAI and APA will convene teens, families, educators and mental health professionals to study how the technology is already appearing in homes, schools and therapy.
Arthur C. Evans Jr., APA's chief executive, said the partnership is aimed at "making sure that knowledge reaches the people building these systems." Sara Johansen, OpenAI's head of mental health and well-being product policy, said AI should reinforce the human relationships and care that young people depend on.
APA moves from reviewer to working partner
The agreement formalizes an existing relationship. APA reviewed an early version of OpenAI's under-18 behavior rules before the company published the principles in December 2025. Those rules direct ChatGPT to use additional caution around self-harm, suicide, sexualized roleplay, dangerous activities, body image and requests to conceal unsafe behavior.
APA's role is now expanding from reviewing a policy document to helping shape resources and safeguards around the product. That gives OpenAI access to a large professional organization with expertise spanning developmental psychology, clinical practice and psychological research. It also ties the partnership to standards APA has already stated publicly.
In its health advisory on generative AI and mental health, APA warned that general-purpose chatbots such as ChatGPT were being used for emotional support despite lacking the clinical training, oversight and regulatory obligations attached to licensed care. APA recommended that chatbots remain subordinate to qualified mental health providers and called for stronger protections for children, teens and other vulnerable users.
The advisory specifically urged developers to address emotional dependency, disclose data practices, connect users with human crisis services and conduct independent predeployment testing for systems available to minors. APA also warned that personalization, constant availability and agreeable responses can amplify existing vulnerabilities or reinforce unhealthy beliefs.
Those positions make APA a consequential partner for OpenAI. The association has already defined a demanding standard for age-appropriate design, and the new partnership moves its experts closer to the teams translating those standards into guidance and safeguards.
OpenAI expands its safety program under pressure
OpenAI has spent the past year adding systems aimed at sensitive conversations and younger users. Its measures include parental controls, safety notifications, age prediction, break reminders and localized crisis resources. In May, OpenAI described a system that carries limited safety context across conversations so ChatGPT can recognize risks that emerge gradually rather than in a single message.
OpenAI says it works with more than 260 mental health experts. Its published performance figures come from internal evaluations and measure behavior against taxonomies developed by OpenAI and outside clinicians, rather than clinical outcomes for users.
The partnership also arrives as mental health safety has become a legal issue for the company. In a February 2026 update, OpenAI said a California court had coordinated multiple mental health-related cases involving ChatGPT into a single proceeding. OpenAI said it would continue improving how ChatGPT identifies distress, de-escalates conversations and directs users toward human support.
APA gives OpenAI a route to turn broad safety commitments into material that parents, schools and practitioners can use. APA, in turn, gains direct access to a company making design choices that already affect how young people seek information and emotional support. The partnership's impact will rest on whether psychological guidance changes model behavior and product decisions, where errors carry greater consequences than a flawed informational answer.