MeditronFO: EPFL launches the first fully transparent medical AI model

As artificial intelligence becomes increasingly prevalent in doctors’ surgeries and hospitals, one question arises: can we trust systems whose inner workings remain opaque?

The LiGHT Laboratory (Laboratory for Intelligent Global Health Technologies) at EPFL , led by Professor Mary-Anne Hartley, has just released MeditronFO, a ‘Fully Open’ medical AI architecture — one that is entirely open and verifiable. Unlike most models presented as transparent, which publish their code without revealing the training data or the methods used, MeditronFO makes the entire process accessible for independent scrutiny by experts, hospitals and regulators. “We would never trust a doctor whose training cannot be verified. The same should apply to AI in healthcare,” explains Xavier Theimer-Lienhard, the PhD student leading the project.

Built on Meditron, a tool developed at EPFL in 2023, MeditronFO also draws on the MOOVE platform, which enables practitioners to contribute directly to the evaluation and continuous improvement of the model, anchoring its development in clinical reality.

The team is now preparing multi-centre clinical trials as part of the MED.USE pilot project, in partnership with the CHUV, to assess in practical terms how doctors interact with AI, whether they follow its recommendations — and what the consequences are for patients. The aim is to demonstrate that AI can improve the quality of care whilst reducing unnecessary procedures.

For Professor Hartley, the aim is clear: healthcare professionals and public authorities must regain control over the AI tools they use, without being forced to rely on proprietary systems whose origins and biases they do not understand.

➡️ Source: EPFL | 📸 ©Canva Media Library