OpenAI AI and Mental Health Grant Program — Closed Historical Reference ($5K–$100K per Project, $2M Total)

OpenAI AI and Mental Health Grant Program — Closed Historical Reference ($5K–$100K per Project, $2M Total)

OpenAI’s AI and Mental Health Grant Program was an announced call for independent research at the intersection of artificial intelligence and mental health. OpenAI announced the program on December 1, 2025, and accepted submissions through December 19, 2025. That application window is closed. OpenAI’s January 28, 2026 update says that it received more than 1,000 submissions, reviewed them, and notified applicants whose proposals were funded. The update says OpenAI is exploring ways to expand and build on the work in the future; it does not announce a second round, a new application window, or a future deadline.

This page is therefore an archive of the completed call. There is no active application to submit through it. The retained deadline below is the original deadline for the closed round; it is not a date for a current or anticipated competition. Details about a possible future program should be checked against a new official announcement because OpenAI has not confirmed whether it will repeat this call or whether any future terms would be the same.

At a glance

What the completed call was designed to fund

OpenAI described the program as support for research proposals that deepen understanding of how AI and mental health overlap, including both potential risks and benefits. The stated aim was to help build a safer and more helpful AI ecosystem. OpenAI said it was particularly interested in interdisciplinary work combining technical researchers with mental health experts or people with lived experience.

The call was framed around independent research, rather than general product development. OpenAI explicitly said it sought to fund research rather than for-profit initiatives and would not prioritize for-profit organizations at that time. That wording describes prioritization; the published eligibility text did not state that every for-profit applicant was categorically barred. A future call could set different terms, so applicants should not carry this preference forward as a current rule without checking a new announcement.

OpenAI named clear deliverables as a useful feature of proposals. Its examples included datasets, evaluations, and rubrics, as well as actionable insights that could inform safety work and the broader AI and mental health community. Examples of insights included synthesized views from people with lived experience, descriptions of how mental health symptoms manifest in a particular culture, and research on language or slang that automated classifiers may miss. These examples indicate the kind of usable research contribution OpenAI described; the page did not present them as an exhaustive or mandatory checklist.

Research areas named by OpenAI

The announcement listed possible topics but explicitly said the examples were not comprehensive. It invited proposals on related directions too. The examples covered distinct research problems:

• How expressions of distress, delusion, or other mental health-related language vary across cultures and languages, and how those differences affect AI detection or interpretation.
• What people with lived experience consider safe, supportive, or harmful when interacting with AI chatbots.
• How mental healthcare providers use AI tools in practice, including what works, what falls short, and where safety risks arise.
• Whether AI systems can promote healthy, pro-social behavior and reduce harm.
• Whether model safeguards remain robust to vernacular language, slang, under-represented linguistic patterns, and low-resource languages.
• How AI systems should adapt tone, style, and framing for youth and adolescents, with possible outputs such as evaluation rubrics, style guidance, or annotated examples.
• How stigma associated with mental illness may appear in model recommendations or interaction styles.
• How systems interpret or respond to visual indicators associated with body dysmorphia or eating disorders, including ethical multimodal datasets and evaluation tasks.
• How AI can provide compassionate support to people experiencing grief, with possible response examples, tone guidance, or evaluation rubrics.

These topics can help explain the program’s scope to researchers reviewing the archive. They are not assurances that OpenAI will fund the same topics in any future grant round. The common thread in the published examples is a research question about system behavior, human experience, evaluation, or safety in a mental health context.

Eligibility and funding terms in the archived round

The official eligibility criteria were brief. Applicants had to be at least 18 years old and be affiliated with a research institution or organization, and/or have significant experience with mental health. The wording allowed the affiliation and experience criteria to be read as alternatives or in combination; the public page did not limit applications to universities or licensed clinicians. It also did not publish a more detailed definition of “significant experience.”

OpenAI said it would not prioritize for-profit organizations at that time because the program sought research rather than for-profit initiatives. This is a prioritization statement, not a universal prohibition in the published criteria. The call invited independent researchers outside OpenAI and highlighted interdisciplinary work, including technical researchers, mental health experts, and people with lived experience.

Proposed budgets were to be between $5,000 and $100,000, with up to $2 million available across the program. The application page stated that reasonable and necessary direct and indirect costs incurred to perform funded activities were allowable, consistent with the recipient institution’s policies. The published information does not provide award amounts for individual recipients, a standard award size, or a breakdown of how the total program funding was divided. The amounts in this archive describe the call’s stated limits, not awards still available.

Application and review process as published

The original application page directed applicants to an online application form. Submissions were open from December 1, 2025, until December 19, 2025, at 11:59 p.m. Pacific Standard Time. OpenAI said that a panel of internal researchers and experts would review applications on a rolling basis and that selected proposals would be notified on or before January 15, 2026. OpenAI also stated that it would not provide individual status updates to applicants whose proposals were not selected.

The form is no longer a route to apply to this completed round. Its continued display of the original opening and closing dates is useful as a record of the call, not evidence that submissions remain available. OpenAI’s January 28, 2026 update subsequently confirmed closure, described the volume of more than 1,000 entries, and said funded applicants had been notified. The same update said the organization was exploring ways to expand and build on the work. It did not set an application procedure, schedule, or eligibility rules for a future round.

The public program materials do not provide a full required-materials checklist or a detailed scoring rubric. They do not specify a universal proposal length, mandatory budget template, CV format, support-letter requirement, or ethics-review document for every applicant. Those materials may be relevant to responsible research planning, but they should not be represented as formal requirements of this call based on the published text. Anyone using this page as a reference should distinguish OpenAI’s stated application process from good-practice preparation advice added by third parties.

How to use this archive

For researchers, the closed call remains a concise record of the questions OpenAI publicly identified as relevant to its safety and well-being work. It can help with retrospective analysis of the program’s research priorities, funding range, stated eligibility, and review approach. It can also help teams decide whether their past proposal aligned with the published scope. It cannot establish what a future funding program will ask applicants to provide.

For organizations planning work in this area, the examples point to the importance OpenAI placed on concrete, useful evidence. A project looking at language variation, for example, might produce documented findings or an evaluation resource that makes failure patterns visible. Work based on lived experience could synthesize those perspectives into actionable insights, while provider research could clarify how tools are used and where safety gaps occur. These are illustrations of the announced interests, not instructions for a new application.

The call’s interdisciplinary emphasis also matters when interpreting its scope. The announcement did not treat mental health as a problem that can be addressed through technical evaluation alone. It specifically mentioned combining technical researchers with mental health expertise or lived experience, and it described outputs that could represent culture, language, and human perspectives. A proposal assessed against this historical call would therefore need to be understood in relation to both AI behavior and mental health context, not solely model capability.

The examples also show why careful evaluation was relevant to the program. Language about distress can differ among cultures and languages; slang and low-resource linguistic patterns can challenge safeguards; and visual signals connected to body image or eating disorders raise questions about ethical data collection and annotation. Youth interactions, stigma, grief, and provider workflows each introduce different contextual considerations. The announcement did not supply a single evaluation method for these problems, but it made clear that useful outputs could include datasets, evaluations, rubrics, taxonomies, papers, and prototype interaction flows.

Current status and official links

As of the latest official update represented here, the application round is closed and no next round has been announced. OpenAI’s statement that it is exploring how to expand or build on the work is not a promise that another grant call will open. Do not use the archived deadline as a future submission date, and do not assume that the earlier amount, eligibility language, or process will carry forward unchanged.

The official OpenAI announcement and status update is the best source for any later news. The original application page records the closed dates and the terms shown to applicants. Questions about the historical program can be directed to [email protected], though the published page does not promise individual application-status updates. If OpenAI publishes a new call, rely on that new official notice for its deadline, amount, eligibility criteria, application steps, and required materials.