Setting the course for European AI sovereignty

Group picture: European experts at the “Pretraining and Posttraining of Sovereign Foundation Models” workshop ahead of the Summit on European Digital Sovereignty
European researchers at the “Pretraining and Posttraining of Sovereign Foundation Models” workshop ahead of the Summit on European Digital Sovereignty. Copyright: Tobias Koch

Berlin, November 17, 2025. Europe is pooling its expertise to significantly advance the development of sovereign AI models. Initiated by the German AI Association and the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, leading experts in German and European AI model development came together at the Digital Technologies Forum in Berlin for a specialist workshop entitled “Pretraining and Posttraining of Sovereign Foundation Models.” As a result, the participants published a joint statement entitled “Joining Forces for AI Development,” which addresses necessary areas of action and calls on business and politics to create long-term, binding, and joint structures to secure Europe’s competitiveness in the field of large AI models. The workshop was actively co-organized by scientists from the Lamarr Institute. The Lamarr Institute is one of the research institutions currently playing a decisive role in developing the technical foundations of sovereign AI in Europe.

The current technical debate centers on the question of how European foundation models can be developed in such a way that they remain powerful and transparent in the long term while also addressing the needs of our industry. The focus is primarily on high-quality data sets, the targeted use of synthetic data, finely tuned curriculum design, and improving the reasoning abilities of the models. These scientific priorities were outlined by Lamarr Innovation Group Leader Mehdi Ali for large language models, who identified the key challenges and starting points from a research perspective and highlighted the factors that are likely to be particularly decisive for the quality and scalability of European foundation models in the future: “European AI development based on high-quality data, open research, and common technical standards creates the foundation for robust and trustworthy models – and thus for true sovereignty in the field of AI,” said Mehdi Ali.

For the Lamarr Institute, one thing is certain: technological sovereignty in Europe can only be achieved through close scientific cooperation, common standards, and coordinated further development of research infrastructure. The workshop preceded the Summit on European Digital Sovereignty, where leading decision-makers discussed the next steps for Europe’s digital independence. This allowed current research perspectives to be directly incorporated into the strategic discussions.

More news