Meta details model access and clinical-data control in $2 million university program

Meta announced $2 million in grants across five U.S. universities on October 28, alongside model access and direct collaboration with its researchers. The named partners are the California Institute of Technology, The Ohio State University College of Medicine, Louisiana State University, Indiana University and UTHealth Houston.

The amount is $2 million across the five schools, not $2 million for each institution. Meta’s detailed research announcement says support varies by institution and project, with funding for selected projects. It describes partnerships already identified by name, rather than presenting a general application process for any university.

For researchers assessing what the announcement offers, that distinction comes before the model list. Access to a model, funding for a project and time working with the provider’s researchers are different resources; the announcement does not give every partner an identical package or publish a complete allocation by project.

The [research details](https://research.meta.ai/blog/supporting‑medical‑scientific‑research) describe first‑party API access to Meta’s frontier Muse models for Caltech’s work in mathematics and physical sciences. Among the named participants is computer scientist Anima Anandkumar, whose project concerns tools for constructing and checking scientific artifacts. These are research plans, not a published finding that the models have solved the underlying scientific problems. Meta describes the healthcare projects differently. It says the participating universities will run models within their own institutional environments, conduct the research independently and retain control of patient and clinical data. According to Meta, that data will not be shared with, processed by or otherwise made available to the company. That is a more specific arrangement than saying a university “uses Meta AI.” The Caltech description identifies a provider API; the healthcare description identifies institution‑run execution and a stated boundary around clinical data. Those descriptions should not be collapsed into a single claim that every project sends its inputs to the same service, or that every partner runs the same software on the same infrastructure.

Meta [says faculty brought their own questions, designed their studies and decided how to use the tools](https://research.meta.ai/blog/supporting‑medical‑scientific‑research). In that account, the universities supply the research design and domain context, while the provider supplies a varying combination of technology, collaboration and funding. Keeping model execution and clinical inputs within an institutional environment does not, by itself, establish the full set of rules governing a project. The announcements do not publish every collaboration agreement, access policy or data‑retention procedure. Meta’s non‑sharing statement is therefore an attributed description of the arrangement, rather than an independent security assessment or a guarantee of regulatory compliance. The governance question is who operates the model and controls the inputs, not simply whose name appears on it. That question remains useful even when several projects share a model family.

The [project descriptions](https://research.meta.ai/blog/supporting‑medical‑scientific‑research) include work at different stages, from planned studies to prototypes and early internal tests. The UTHealth Houston discussion explicitly separates initial workflow validation from clinical outcome measures. The existence of funding or a working prototype should not be presented as proof of improved care. The current development is a set of named research collaborations with distinct access and data arrangements. It does not announce a broadly available diagnostic product or show that the projects have delivered the medical and scientific benefits described as their goals. Those outcomes require evidence separate from the partnership announcement. Anima Anandkumar speaking at the AAAS Annual Meeting on February 14, 2026. Meta identifies her among Caltech’s research participants. Archive photograph, not the October partnership announcement or evidence of clinical validation. Photo: Runawaymo, via [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Anima_Anandkumar_speaking_at_AAAS_2026.jpg), [CC BY‑SA 4.0](https://creativecommons.org/licenses/by‑sa/4.0/). Photograph cropped and downsampled; no artificial upscaling or added margins. Image adaptation retains the same licence.