Five Million Dollars for Teen AI Research: OpenAI Opens Its Most Self-Critical Grant Program Yet

Earlier this month, OpenAI quietly opened applications for a grant program that may be the most self-critical thing the company has funded this year. The AI and Teen Development Research Grant Program will commit up to $5 million to independent researchers studying how generative AI affects the lives and development of young people aged 13 to 17, with a stated focus on teens’ social and emotional development. Submissions opened on September 8, 2026 and close on October 6, 2026 at 11:59 PM PDT, with selected proposals notified on or before November 13.

On its face, this looks like routine corporate philanthropy. It isn’t. The program arrives at a moment when OpenAI is deliberately steering teenagers into its products — and when the outside evidence about what that does to them is thin, contested, and increasingly alarming.

What the program actually funds

According to the official application portal, OpenAI will award individual grants of up to $1 million, totaling up to $5 million across the cohort. A maximum of 10% of any project’s budget can go to overhead or indirect costs — an unusually tight cap that pushes money toward actual research rather than university administration.

The scope is deliberately broad. OpenAI says it welcomes proposals that deepen understanding of:

• How teens actually use generative AI, and how different patterns of use relate to developmental outcomes
• Potential impacts of AI use in teens’ lives — positive and negative
• The individual, social, cultural, and technical factors that shape those effects
• Technical interventions, safeguards, and design choices that could increase benefits or keep teens safer
• How effects differ across ages, developmental stages, populations, cultures, languages, and socioeconomic contexts

That last bullet is the tell. OpenAI isn’t just asking whether AI is “bad for teens” — it’s asking for granular, mechanism-level evidence about which teens, in which contexts, under which conditions. Qualitative, quantitative, experimental, observational, theoretical, participatory, and mixed-methods approaches are all eligible. Submissions are open to researchers from any country, with a stated preference for regions where AI use among teens is high.

Eligibility requires applicants to be 18 or older and affiliated with a research institution, or to have significant relevant experience in child or adolescent development, well-being, human-computer interaction, AI impacts, or adjacent fields. For-profit initiatives will not be prioritized — this is research funding, administered by OpenAI Group PBC, not the OpenAI Foundation.

The context the company can’t ignore

Why is OpenAI spending $5 million to study its own product’s effects on teenagers? Because it has spent the past year making teenagers a core audience.

In August 2026, the company launched ChatGPT for Teens, an age-aware experience that automatically routes users OpenAI believes to be 13–17 into a protected environment with stricter content guardrails, Study Hours, and expanded parental controls. The launch made teen protection the default rather than an opt-in — a genuinely defensive design choice, but also a strategic one: it positioned ChatGPT as a legitimate educational tool for minors, not just a consumer chatbot that happens to be used by them.

The problem is that product strategy has outrun the evidence base. The most cited statistic in this space comes from Common Sense Media’s July 2025 national survey, which found that 72% of American teens — nearly three in four — have used an AI companion, with about half using one regularly and one in three using AI companions for social interaction, including romantic and emotional conversations. Common Sense’s review of the evidence was blunt enough that the organization recommended no one under 18 use AI companions at all.

That recommendation and OpenAI’s teen strategy are on a direct collision course. Regulators in multiple jurisdictions are actively debating age-appropriate design codes, parental consent requirements, and companion-bot restrictions for minors. OpenAI’s own framing acknowledges the stakes: the company says it wants “a stronger evidence base about the way teens use AI… to inform development of AI products for teens and to help answer major questions facing policymakers and regulators.”

Read that sentence carefully. OpenAI is explicitly funding research it expects to be cited in product decisions and regulatory proceedings — including proceedings that could go against it.

The independence question

Any industry-funded research program invites an obvious question: who controls the findings?

The structural details matter here. Applications are reviewed by “a panel of internal researchers and experts” on a rolling basis, and OpenAI notes it will not provide status updates for unsuccessful applications. The program is funded and administered by OpenAI Group PBC — the public-benefit corporation entity, not a legally separate foundation. Researchers taking this money will be studying the effects of products made by the entity paying their invoices.

Skeptics will point to the pattern: technology companies funding safety research that reliably finds its way into policy arguments favorable to those companies. Optimists will point to the alternative — no funded research at all, and product decisions made on vibes. The program’s genuinely encouraging signals are the breadth of eligible methodology (participatory research, for instance, means teens themselves shaping the studies), the any-country eligibility, and the explicit openness to findings about harm.

There’s also a defensive reading, and it’s not wrong: $5 million is roughly rounding error against OpenAI’s operating budget, and producing an evidence base — even an ambivalent one — gives the company something concrete to point to when regulators, plaintiffs’ lawyers, and parents come asking what it knew and when. The BIPA class action over Meta’s Nametag weights, currently working through US courts on precisely a “what did the training data do” theory, is a live demonstration of how expensive that question can become.

Why the deadline matters

The timeline is compressed in a way that suggests urgency. A four-week application window (September 8 to October 6), rolling review, and notification by November 13 means funded research could begin before the end of the year — fast by academic standards, where grant cycles typically run six months to a year.

That speed is interpretable in two directions. Charitably, OpenAI wants answers before its teen user base grows further, and knows the research takes years once started. Less charitably, it wants citable interim findings in hand for the policy fights of 2027.

Either way, the practical advice for researchers is simple: this is a live, well-funded, unusually broad call for proposals in a field where funding is scarce. The contact address is collaborativeresearch@openai.com, and applications run through OpenAI’s portal on SurveyMonkey Apply.

The bigger picture

The uncomfortable truth underneath this program is that the world’s most valuable AI company is running a live experiment on its youngest users and doesn’t know the results. Teens are adopting conversational AI faster than any studied population, the products are being explicitly designed and marketed to them, and the developmental literature is years behind deployment. OpenAI funding the study of its own products’ effects on adolescents is better than nobody funding it — but it’s also an admission that the evidence gap was allowed to open in the first place.

Five million dollars won’t close it. A cohort of independent, methodologically diverse studies on teen AI use — with real budgets, real timelines, and publication rights that survive OpenAI’s review — would at least be a start. Watch what the selected proposals look like in November. That roster will tell you more about OpenAI’s actual appetite for uncomfortable findings than any safety blog post.

For now, the clock is running: applications for the AI and Teen Development Research Grant Program close October 6, 2026 at 11:59 PM PDT.