Meta models enter real research at five universities

Meta AI Research announced projects with Caltech, Ohio State, Louisiana State University, Indiana University and UTHealth Houston. Universities will run Meta models inside their own environments for healthcare work, retain control of patient and clinical data, and not share that data with Meta. Projects include combining abdominal CT, records and ECG data with Muse to identify HFpEF risk earlier; a clinician-in-the-loop multi-agent system for Alzheimer's diagnosis; bedside and edge use of SAM to detect infant tracheostomy displacement; offline multilingual newborn-care guidance on low-cost Android devices using SeamlessM4T; and Lean 4 tooling that makes model-generated scientific artifacts formally checkable. Some efforts are prototypes, retrospective cohorts or planned user studies rather than clinical products. Accuracy, bias, clinician reliance, privacy governance and regulation still need independent validation.