A LANDMARK initiative has begun to create the world’s first truly global medical AI foundation model, bringing together data and expertise from over 100 research groups across 65 countries. The Global RETFound Consortium aims to build a vast AI system trained on more than 100 million eye images, producing a geographically and ethnically diverse dataset for medical applications.
Medical artificial intelligence has enormous potential for disease detection and personalised healthcare, but most models so far have been limited by narrow geographic and demographic data. This has led to significant gaps in the generalisability of AI tools, especially in under-represented regions. Addressing these shortcomings, the Global RETFound project is pooling fundus photographs from every continent except Antarctica, allowing much broader participation while upholding strict data privacy standards.
Led by institutions including the National University of Singapore Yong Loo Lin School of Medicine, Moorfields NHS Foundation Trust, University College London, and the Chinese University of Hong Kong, the consortium uses a dual framework for data sharing. Some groups fine-tune generative AI models locally and share only model weights, keeping patient data secure, while others contribute de-identified data via a secure infrastructure. This approach ensures collaboration is open to institutions with varying resources and technical capabilities. The assembled model will be released under a Creative Commons licence for non-commercial research, with comprehensive evaluation in diseases such as diabetic retinopathy, glaucoma, age-related macular degeneration and cardiovascular conditions.
For clinical practice, this initiative could set new standards for fairness and global generalisability in medical AI, accelerating the development of tailored diagnostic tools that serve local needs more effectively. As the methodologies extend beyond ophthalmology, similar international collaborations may emerge in other specialties. Researchers worldwide are invited to join, with the goal of creating equitable and actionable AI systems for healthcare.
Reference
Tham YC et al. Building the world’s first truly global medical foundation model, Nature Medicine. 2025;DOI:10.1038/s41591-025-03859-5.
