OpenAI $5 million teen AI research grants

OpenAI is putting OpenAI $5 million teen AI research grants on the table for independent studies of how generative AI shapes 13 to 17 year olds, according to OpenAI News.

Applications opened September 8 2026 and close October 6 2026, with decisions by November 13 and interim updates due in Q1 2027.

Why this matters for AI and startups

For builders shipping teen-facing products, the evidence gap is now a liability.

Regulators in the US, UK and EU are pressing labs on whether chatbots affect emotion regulation, identity, agency and human connection, and product teams have little peer-reviewed data to answer.

OpenAI frames the program around four priorities that map to product requirements: emotional well-being, social development and relationships, variance by time, use type and demographics, and efficacy of mitigations and age appropriate design.

Those questions will soon show up in trust and safety reviews, procurement checklists and policy filings.

Each grant caps at $1 million within a $5 million total pool, with overhead limited to 10 percent, which favors focused empirical work over large institutional overhead.

Eligibility is broad: any researcher 18 or older affiliated with a research institution or with significant relevant experience, with applications in English but research in any language.

OpenAI says it will prioritize interdisciplinary teams, underrepresented populations or regions, countries with high AI adoption, and projects centered on ChatGPT that can inform near term policy.

Grants are funded and administered by OpenAI Group PBC, not the Foundation, and final awards require a grant agreement and compliance with its terms.

What the source misses

The announcement says little about independence, data access and what counts as success.

Publication in a peer-reviewed journal or public report is strongly encouraged but not a condition of funding, so findings can end as a private report to OpenAI.

Review is by a panel of internal experts and advisors on a rolling basis, with no named external chair, scoring rubric or commitment to publish all results.

There is no offer of privacy-preserving access to ChatGPT teen usage data, model behavior or evaluation tooling, so grantees must recruit and measure independently.

That raises feasibility questions for studying safeguards or usage patterns at scale without privileged data.

Projects involving sensitive personal, behavioral or mental health information remain eligible but need additional ethics review, safeguarding expertise and time that sits uneasily against the Q1 2027 interim update.

For startups, the takeaway is practical: fundable evidence on teen safety will soon be an expectation, not a differentiator.