OpenAI Is Picking the Outside Researchers Who Will Study How Its AI Affects Teens

OpenAI is putting $5 million behind outside research on how its generative AI products affect 13- to 17-year-olds. The company whose products are being studied is also the one choosing the researchers, and the public announcement does not show the mechanism that would make the work independent. (OpenAI announcement)

The program, as described in the OpenAI post, funds outside research on how generative AI affects adolescent development, with a stated focus on social and emotional development during the teen years. OpenAI says the work is meant to "inform product development" and to "answer questions facing policymakers and regulators." Those are the company's stated motivations for the program, not verified claims about its effect.

The five research areas are wide. They cover how teens actually use generative AI and how those patterns relate to developmental outcomes; potential impacts on teens' lives; the individual, social, cultural, and technical factors that shape those effects; technical interventions, safeguards, and design choices that could increase benefits or reduce harm; and how the effects differ across ages, developmental stages, populations, cultures, languages, and socioeconomic contexts. Eligible methods are listed broadly: qualitative, quantitative, experimental, observational, theoretical, participatory, and mixed-methods. The interdisciplinary scope runs from child and adolescent development and psychology to public health, HCI (the field that studies how people interact with computer systems), communication, sociology, anthropology, data science, and AI safety. Proposals are open from any country, with prioritization for countries with significant AI adoption.

The announcement leaves out the load-bearing part: the review panel, the conflict-of-interest posture, the data ownership terms, and the publication clause. It does not say whether grantees can publish findings OpenAI disagrees with. It does not name a teen advisory body for the program itself, even though the subject is adolescent development. It does not link to a request for proposals, an application deadline, or award sizes. The visible excerpt of the post is truncated mid-sentence at "submissions that relate to one or m" — the application mechanics, including the RFP link itself, are not in the public text a reader can see today.

Five things are worth tracking over the next 12 to 24 months, as the funded portfolio takes shape:

1. Which questions got funded. The announcement names five areas. The funded share will tell readers what OpenAI considers in-scope and, by omission, what is not.
2. Which got passed over. The list does not cover product-roadmap influence on teen experience, internal data access, or post-deployment harm. The gaps in the funded set are part of the story.
3. Whether teens are part of the design. The post invites participatory research with teens but does not name a teen advisory body for the program.
4. Whether findings can be published regardless of result. A corporate funder's publication clause is the test of independence in practice.
5. Whether pre-registration is required. Pre-registration means committing to the study design and outcome measures before data is collected. It is the standard guard against results-driven reframing.

The re-reports that followed — at InsideAI, Blockchain News, and StartupHub.ai — added no facts beyond the OpenAI post. They confirm the announcement is being circulated. They do not confirm the program is structured to produce an independent evidence base.

The dollar figure is real. The review panel, the publication clause, and the data ownership terms are not yet public. When those appear, they will tell readers more than the announcement does.