OpenAI announced a collaboration with the American Psychological Association (APA) focused on the use of artificial intelligence by young people, with special attention to well-being and mental health. The most important point to interpret the news is to separate what already works in ChatGPT from what the partnership intends to develop from now on.
There is also a relevant date correction: although this article is published on August 7, 2026, OpenAI's official statement is dated August 6, 2026. The text presents collaboration as a way to incorporate psychological science into the development and responsible use of AI among young people.
The advertisement does not present results of clinical trials, does not demonstrate that ChatGPT is effective as a psychological treatment and does not transform the APA into a clinical certifier of the product. An important part of the deliveries described is still in the planning stage.
What OpenAI and APA announced
According to OpenAI, the collaboration should involve families, health professionals, school psychologists, educators and young people themselves. The company says it intends to work with the APA to better understand how teenagers use AI, where difficulties arise in schools, families and therapy settings, and in which situations technology could responsibly help with information search or referral to human support.
Among the fronts announced are the production of resources aimed at parents and guardians, materials for clinicians and school psychologists and new conversations with teenagers, families and professionals. The proposal is also to use this learning to guide future safeguards and evidence-based approaches.
These formulations matter because they indicate work objectives, not results already achieved. The statement uses expressions such as “we plan to develop”, “planned work” and “over time”, without providing a public timeline for all deliveries, quantitative safety targets or expected clinical results.
Collaboration also doesn't start from scratch. OpenAI reports that the organizations had already held a meeting with researchers, clinicians, educators, mental health organizations and youth representatives. Additionally, the APA had previously participated in reviewing a draft version of the OpenAI principles for users under the age of 18.
What already exists and what is still a promise
| Measure or initiative | Situation on August 7, 2026 |
|---|---|
| U18 Principles for Model Behavior | They already exist. They were published in December 2025 and describe additional protections aimed at teenagers aged 13 to 17. APA participated in reviewing a draft. |
| Parental controls in ChatGPT | Already exist. Accounts of guardians and teenagers can be linked; Those responsible can manage certain settings, define break times and receive alerts in specific situations. The feature does not give access to the teenager's conversations. |
| Safety notifications for guardians | Already exist in linked accounts. Can be sent in limited situations after detecting and reviewing signs related to, for example, serious self-harm concerns. |
| Age prediction | Under implementation. OpenAI documentation states that the system is being rolled out globally to help apply expertise aimed at under-18s. This does not mean identical availability in all markets. |
| Localized crisis resources | Already introduced. OpenAI documents localized resources and easy access to helplines through partnership with ThroughLine. |
| New materials for parents and guardians developed with APA | Planned. The announcement does not provide a release date for the announced set. |
| Resources for clinicians and school psychologists | Planned. Must address healthy use of AI, over-reliance, and identification of problematic patterns. |
| New consultations with young people, families and professionals | Envisaged as part of the partnership. The stated objective is to understand real uses, gaps and needs before guiding new measures. |
This division avoids assigning resources that were previously released to the partnership. Parental controls, U18 principles, alerts and age prediction are part of a security strategy that was already underway before the August announcement.
The partnership is not a clinical validation of ChatGPT
The participation of the American Psychological Association may give the work greater access to knowledge in psychology and human development, but should not be interpreted as proof that a general-purpose chatbot can replace psychotherapy or professional care.
The APA itself takes a cautious stance. In its health advice on generative AI chatbots and wellness apps, the association states that these technologies should not replace a qualified mental health professional. The entity also notes that many of these systems were not developed to provide clinical treatment, may not have sufficient scientific validation, and require specific safeguards for children, adolescents and other vulnerable groups.
In other guidance dedicated to AI and adolescent well-being, the APA highlights that research into the effects of these technologies is still developing and that the impacts cannot be classified simply as positive or negative. Among its recommendations are AI literacy, critical evaluation of the responses generated, and protections appropriate to the developmental stage of young users.
Therefore, phrases like “the partnership proves that AI improves the mental health of adolescents” or “the APA validated ChatGPT as a therapeutic tool” would go beyond the available evidence.
What still needs to be observed
The August announcement sets a direction, but still leaves questions that can only be assessed as initiatives are published and implemented.
It will be necessary to check which materials will be produced in conjunction with APA, when they will be available, for which audiences and countries and how they will be updated. It will also be relevant to observe whether future changes to the product will have public evaluation criteria, specific tests for adolescents and evidence that allows measuring their real effects.
The same care applies to expressions such as “support in times of suffering”. The statement says that the partnership intends to study how AI can help young people in these situations, but this does not equate to stating that a chatbot has demonstrated therapeutic efficacy. OpenAI itself presents crisis resources as ways to connect users to human support, not as a substitute for emergency or mental health services.
For now, the concrete news is institutional: OpenAI and APA are expanding a collaboration that already had a history and defined new work fronts. Part of the protections mentioned by OpenAI can already be documented in the product; new joint capabilities, their effects, and their ability to reduce risk need to be evaluated as they are implemented.
