Several European countries are developing their own artificial intelligence systems to reduce reliance on American tech companies, three years after ChatGPT brought AI into mainstream use. Known as sovereign AI, these systems are built, hosted, and governed domestically to serve local populations rather than relying on foreign platforms or cloud services.
The European Parliament highlighted in a June report that the continent is heavily dependent on foreign technologies, particularly from the United States. The report noted that the U.S.’s $500 billion domestic investment in AI is likely to maintain this technological lead. The EU said it must invest in research and develop domestic AI systems, with national governments taking a central role.
Germany recently launched its Sovereign Open Source Foundation Models (SOOFI) project, aiming to create an advanced open-source AI model adaptable for various commercial applications. The system is intended for complex tasks such as AI-controlled robotics. Wolfgang Nejdl, a professor at Leibniz University Hannover involved in the project, said SOOFI will ensure Europe has powerful AI tools aligned with European values, crucial for education, medicine, administration, and production. Deutsche Telekom and T-systems are supporting the project, which plans to use 130 NVIDIA chips and over 1,000 GPUs to train a model with 100 billion parameters.
Switzerland introduced Apertus, the country’s first multilingual AI model, trained on 15 trillion data points across more than 1,000 languages, including Swiss German and Romansh. Developed by ETH Zurich and collaborators, Apertus is fully open-source, allowing researchers and the public to customize it for areas such as law, climate, health, and education.
Poland’s Polish Large Language Model (PLLuM) focuses on the specificities of the Polish language, supporting applications from writing assistance to educational tools. The model will eventually integrate into the government’s public administration system as part of Hive AI, providing virtual assistants for citizens and classrooms.
Spain has launched Alia through the Barcelona Supercomputing Centre, offering a multilingual AI infrastructure in Spanish, Catalan, Basque, and Galician. Alia aims to develop responsible AI for public services, including health diagnostics and tax support. Earlier initiatives like Catalonia’s Aina project have also created AI models for regional languages.
In the Netherlands, three non-profits are developing GPT-NL, a Dutch-language open-source AI model trained on a mix of licensed, public, and synthetic data. Portuguese universities have created Amalia, a Portuguese-language AI designed for public administration and scientific analysis, with a beta test underway and a planned public release in mid-2026.
These initiatives reflect Europe’s growing focus on creating domestic AI infrastructure, ensuring technological independence while promoting accessibility, local language support, and ethical standards.
