Federated learning in healthcare: Addressing AI challenges and operational realities under the GDPR
Federico Vota, Francesca Pediconi, Alessandro Liscio · Journal of data protection & privacy. · 2025
The fundamental characteristic of machine learning (ML) algorithms is their ability to learn to solve problems autonomously, based solely on the data provided to them; to do so, ML models require a huge amount of data (ie training data) to learn how to solve the problems they are subjected to. When talking about the training of artificial intelligence (AI) algorithms, especially for healthcare use, the matter of personal data included in the training datasets cannot be ignored. This includes data belonging to ‘special categories’ (including ‘health data’), which require more robust measures than those put in place for processing of so-called ‘common data’ in order to be processed under the General Data Protection Regulation (GDPR). Among those in the healthcare sector who develop or use AI, this issue is highly relevant. Federated learning is a cooperative ML technique capable of exploiting the knowledge stored in multiple datasets without the need to pool them out, as an innovative approach to privacy-preserving AI model training. In particular, in the healthcare sector, this technique can be used to enable multiple organisations to collaborate without sharing sensitive patient data. The aim of this paper is to showcase how federated learning integrates seamlessly with existing healthcare IT systems to address privacy concerns by presenting use cases that offer a concrete perspective on federated learning’s potential and operational challenge.