The European Commission has awarded the Strategic Technologies for Europe (STEP) award to the multilingual artificial intelligence project OpenEuroLLM. This is an alliance that is creating the first family of large, open-source language models covering all EU languages. What is its goal and prospects?
The European Commission has awarded the Strategic Technologies for Europe (STEP) award to the multilingual artificial intelligence project OpenEuroLLM. This is an alliance that is creating the first family of large, open-source language models covering all EU languages. What is its goal and prospects?
Previously, experts expressed concern that the European Union is losing the race in the development of artificial intelligence. The USA has a number of successful companies, both focused solely on AI (OpenAI, Anthropic) and large corporations with AI divisions (Google, Meta, Microsoft). In China, Alibaba, in particular, is working on AI, and recently everyone has been talking about the successes of the company DeepSeek. In contrast, the EU does not have such prominent companies, but this may change.
Like DeepSeek, the OpenEuroLLM alliance aims to develop next-generation open-source language models, but its goal is quite different: to create a European artificial intelligence that will contribute to the development of digital leaders and effective public services across the continent, The Next Web reports.
To support these goals, OpenEuroLLM is building a family of high-performance multilingual large-scale language fundamental models. The models will be available to commercial, industrial, and government services.
It is an alliance of more than 20 leading European research institutes, companies and high-performance computing centers. It is led by Jan Gajic, a renowned computational linguist from Charles University in the Czech Republic, and Peter Sarlin, co-founder of Silo AI, the largest private artificial intelligence lab in Europe, which was acquired by American chipmaker AMD for $665 million last year.
It also includes a number of promising tech startups. Among them are Aleph Alpha, a leading German AI company, Finland’s CSC, home to one of the world’s most powerful supercomputers, and France’s Lights On, which recently became Europe’s first publicly traded company with GenAI shares.
In 2023, it was reported that Aleph Alpha was working on a potential European response to the American chatbots Google Bard and OpenAI’s ChatGPT. The European chatbot was called Luminous.
OpenEuroLLM is supported by the European Commission. According to Sarlin, this initiative could become the largest AI project in the history of the European Commission.
«The uniqueness of this initiative is that we are bringing together many of Europe’s leading AI organizations in one focused project, rather than many small, disparate projects. This focused approach is what Europe needs to create open European AI models that will ultimately enable innovation at scale,» he said.
He clarified that the OpenEuroLLM budget is €52 million, with computing commitments that could have a higher monetary value. In addition to funding from the Commission, OpenEuroLLM has received support from STE P, the EU’s program to stimulate investment in strategic technologies.
STEP is a new EU initiative aimed at increasing the competitiveness of European industry by supporting critical technologies, including artificial intelligence.
While China and the US are developing new AI capabilities at breakneck speed, Europe faces an uncertain future in the digital landscape.
OpenEuroLLM hopes to strengthen the EU’s position through new digital infrastructure. The project is also committed to implementing AI with European values of democracy, transparency, openness and public engagement.
According to OpenEuroLLM, the models, software, data and assessment will be fully open. They will also be capable of fine-tuning and coaching for the specific needs of industry and the public sector. In addition, the alliance promises to preserve linguistic and cultural diversity.
«It’s not about building a general-purpose chatbot, but about building a digital and AI infrastructure that will enable European companies to innovate with AI. Whether it’s a healthcare company developing specialized assistants for doctors or a bank creating personalized financial services, they need AI models that are adapted to the context in which they operate and that they can control and own,» said Peter Sarlin.
According to him, the project provides European companies with tools to create models and solutions in their languages, which they speak and which they can control.
Recall that one of the European AI projects recently got into trouble. An artificial intelligence chatbot, created with the support of the French government, was shut down just three days after its release after it gave nonsensical answers to simple mathematical equations and recommended that one user try «edible eggs produced by cows.» The model was named after the Australopithecus Lucy.
