Fraunhofer Institute for Systems and Innovation Research ISI
Dr. Bernd Beckert gave a presentation at the IEEE’s “International Conference on Federated Learning Technologies and Applications (FLTA25)”, which took place in Dubrovnik, Croatia, in October 2025.
Federated Learning (FL) is a method for training AI models in which data are not moved to a central server but remain on distributed devices or servers. FL is considered a promising technology because it enables specialized AI models (that are less prone to hallucinations) to be realized based on data that are currently not accessible to large language models. However, its market prospects are still unclear despite the recent significant growth in the number of publications in this field and its large developer community.
Dr. Bernd Becker presented the current status of this new technology’s implementation in applications in his conference paper. Summary: Despite its increasing importance, there are currently few larger, convincing use cases; it has been proven that the frameworks function (technically), but not their practical feasibility. In the discussion of his presentation, calls were made for more transfer projects and an analysis of the federated learning projects already conducted in the areas of health, finance, and production.