Is Your Country Falling Behind? Discover the Secret European Nations Racing to Build Sovereign AI!

Three years after OpenAI launched ChatGPT and thrust artificial intelligence (AI) into the mainstream conversation, European countries are beginning to develop their own sovereign AI systems. This term refers to a nation’s ability to create, host, deploy, and govern AI technologies tailored to its citizens, reducing reliance on foreign, particularly American, systems.

The European Parliament recently acknowledged its heavy dependence on foreign technologies, particularly those emerging from the United States. A June report highlighted that this reliance hampers the EU from nurturing its own tech champions. The report pointed out that as the U.S. invests a staggering $500 billion (€432.9 billion) in domestic AI development, Europe’s dependency appears poised to continue.

To regain its technological edge, the EU has outlined the need for robust investments in research and the development of new systems. National governments across Europe are stepping up to the challenge, with several countries crafting their own sovereign AI initiatives.

The German Initiative: SOOFI

Leading the way is Germany, which recently announced its own AI plan, called the Sovereign Open Source Foundation Models (SOOFI). This initiative aims to build an advanced AI open-source model that other companies can adapt for their AI products. The German government stated that SOOFI will tackle complex tasks like AI-controlled robotics.

Wolfgang Nejdl, a professor at Leibniz University Hannover, remarked, “With SOOFI, we are laying the foundation for the next generation of European AI models – sovereign, powerful, and entirely in European hands.” He emphasized that large AI models that align with European values are vital for fostering trust in AI, especially in sensitive sectors like education, medicine, and administration. Major telecommunications player Deutsche Telekom and its subsidiary T-Systems are providing technical support, aiming for SOOFI to encompass 100 billion parameters that regulate the model’s behavior. To facilitate this, Deutsche Telekom is set to deploy about 130 NVIDIA chips and over 1,000 graphics processing units (GPUs) by next March.

Switzerland's Multilingual Model: Apertus

In September, Switzerland made strides with its launch of Apertus, the nation’s first multilingual language model. The name, meaning “open” in Latin, signifies the model's adaptable nature, allowing researchers, professionals, and the public to customize it for specific needs.

Collaborating with ETH Zurich, Apertus was trained on an impressive 15 trillion tokens across more than 1,000 languages, including Swiss German and Romansh. The developers have ensured that all aspects, from training architecture to datasets, are open for public use. “This release isn't a final step—it's a beginning,” stated Antoine Bosselut, co-lead of the Swiss AI initiative, emphasizing a long-term commitment to “sovereign, open AI foundations that serve the public good, worldwide.”

Poland's Tailored Solution: PLLuM

Poland also unveiled its own domestic large language model, Polish Large Language Model (PLLuM), in February. Specifically designed to accommodate the nuances of the Polish language, the PLLuM aims to address the complexities of inflection and syntax in various AI applications.

Deputy Minister of Digital Affairs Dariusz Standerski highlighted PLLuM as “an investment in the digital state,” asserting that it will evolve into Hive AI, integrating seamlessly into government operations. This system is expected to provide citizens with virtual assistants for public information and automate tasks like document processing.

Spain's Alia Initiative

Meanwhile, in January, Spain's Barcelona Supercomputing Centre launched Alia, the first European open and multilingual AI infrastructure. Developed with the powerful MareNostrum 5 supercomputer, Alia is designed to promote “responsible AI for the service of the people.”

Alia offers a database of resources and tools in multiple languages, fostering the development of domestic models. The Spanish Agency for the Supervision of Artificial Intelligence (AESIA) plans to expand Alia into applications like tax agency chatbots and health diagnostics.

Other National Initiatives

Several other European nations are also advancing their AI capabilities. The Netherlands is working on GPT-NL, an open-source AI model tailored to Dutch culture, with a consortium securing agreements with local publishers for training data.

In Portugal, academics have initiated Amalia, an AI capable of generating code and interpreting information in Portuguese, with plans for public release by 2026. These sovereign AI developments underscore a transformative shift in how Europe aims to regain its digital sovereignty amid the global tech landscape dominated by American firms.

As Europe forges ahead with these initiatives, the implications are profound. National AI systems promise not only to enhance local governance and public services but also to ensure that technological advancements reflect the values and needs of European citizens. The race for AI sovereignty is not just about technology; it is about autonomy, trust, and the future of digital identity in an interconnected world.