Key Takeaways

  • OpenAI is offering up to $5 million for independent research into how generative AI affects people aged 13 to 17.
  • Individual awards can reach $1 million, with applications closing October 6, 2026, and decisions expected by November 13.
  • The program arrives as AI chatbot use for mental health advice grows, raising questions about safety, efficacy, privacy, and adolescent development.

OpenAI has opened applications for a $5 million research grant program focused on how generative AI influences the lives and development of teenagers aged 13 to 17. Individual grants can be worth up to $1 million, giving selected researchers room to conduct work that goes beyond small pilot studies or short-term product testing.

Applications opened September 8, 2026, and will close October 6, 2026. OpenAI plans to make funding decisions by November 13, while grant recipients will be expected to provide interim updates in Q1 2027. The compressed timetable suggests OpenAI wants research projects to begin while teen adoption of generative AI is still evolving rapidly.

The emphasis on independent studies matters. Much of the available evidence about young people and generative AI remains fragmented across surveys, controlled trials, observational research, and product-specific evaluations. Teenagers may use the same chatbot for homework, social advice, entertainment, creative expression, or emotional support, making it difficult to separate benefits from risks.

Generative AI is already thoroughly integrated into the daily routines of many young people. A nationally representative US survey conducted by RAND, Brown University and Harvard researchers found that about 1 in 5 adolescents and young adults had used an AI chatbot for mental health advice by 2025. That was up from 13.1%, or roughly 1 in 8, in a 2024 survey.

Those findings add urgency to questions that enterprise AI buyers will recognize. How should a system respond when a user appears distressed? What information should be retained? When should an automated conversation point someone toward a parent, clinician, counselor, or crisis service? For teenagers, those product and governance questions become more sensitive because age, maturity, family context, and developmental stage can affect both use and outcomes.

Evidence of potential benefit is emerging, although it is not uniform. A 2025 systematic review published by the Journal of Medical Internet Research examined 31 randomized controlled trials involving 29,637 adolescents and young adults. It reported small-to-moderate reductions in mental distress, with a standardized mean difference of -0.35 and stronger effects for depression and anxiety.

That said, positive average effects do not settle questions about long-term reliance, inappropriate responses, persuasive design, or uneven performance across demographic groups. Research focused specifically on 13 to 17 year olds could help developers understand where general-purpose models behave differently from products designed for education, wellbeing, or clinical support.

Specialized vendors are already deploying applications for young users. Mental Health Innovations uses AI-based signposting to connect people with crisis support, while Telefono Azzurro operates AzzurroChat, an AI mental health platform for teenagers. OpenAI has also supported youth-related work through its EMEA Youth & Wellbeing Grant. These initiatives reflect growing demand, but they also create pressure for clearer evaluation methods and age-appropriate safeguards.

For technology vendors, the grant program could eventually influence product design in several areas: age assurance, parental controls, escalation procedures, data retention, model evaluation, and disclosures explaining that a chatbot is not a human adviser. Schools, healthcare providers, nonprofits, and regulators may also gain evidence for procurement and oversight decisions.

There is a strategic dimension for OpenAI as well. Funding external researchers can broaden the evidence base around generative AI while giving the company earlier visibility into concerns likely to shape regulation and customer expectations. Independence will therefore be closely watched. Research design, publication rights, methodological transparency, and the handling of unfavorable findings will affect how much confidence stakeholders place in the resulting work.

The first meaningful signals should arrive through interim updates in Q1 2027. Strong studies could help the industry move from broad claims about teen safety toward measurable outcomes, clearer risk thresholds, and more defensible design choices. For businesses building AI experiences used by young people, that evidence may become increasingly difficult to treat as optional.