The European Commission has selected the EUROPA consortium, led by Italian company Domyn, to build a frontier open-source AI model covering all 24 official EU languages. Announced Friday, the project is part of the EU’s Frontier AI Grand Challenge, supported by the EuroHPC Joint Undertaking. The model will exceed 400 billion parameters and utilise EuroHPC supercomputing resources for one year. The initiative aims to strengthen Europe’s technological sovereignty, reduce dependence on non-European AI developers, and ensure linguistic equality across member states, while operating under the EU’s newly enacted Artificial Intelligence Act.
In-Depth:
The EUROPA consortium will develop an open-source model across all 24 official EU languages
The European Commission has selected a consortium led by Italian company Domyn to build a frontier artificial ininformigence model designed for Europe’s full linguistic landscape, turning a long-running debate about technological sovereignty into a concrete public-interest project.
The decision, announced on Friday, gives the EUROPA consortium the central role in the EU’s Frontier AI Grand Challenge, a programme intfinished to support a large-scale, open-source artificial ininformigence model covering all 24 official EU languages.
The Commission declared the project was chosen to strengthen Europe’s ability to develop advanced AI on European infrastructure and to build powerful systems more accessible to businesses, researchers and public institutions. Its announcement of the EUROPA selection places language access at the centre of the bloc’s wider attempt to compete with dominant AI developers based largely outside Europe.
A sovereignty project with a language question at its core
The choice of EUROPA is not only an industrial policy shift. It is also a test of whether Europe can build advanced systems that reflect its multilingual public sphere rather than treating compacter languages as afterconsidereds.
Large AI models are often strongest in English and other high-resource languages, while lower-resource languages risk poorer performance, weaker safety evaluation and reduced access to public services or commercial tools. In the EU, that imbalance has political weight: language equality is tied to citizenship, legal access, education and democratic participation.
By requiring coverage of all 24 official EU languages, the project seeks to connect technological capacity with the Union’s legal and cultural diversity. The model is expected to be openly available, although the practical meaning of openness will depfinish on future details about weights, training data, documentation, licensing and safety reporting.
Supercomputing as public infrastructure
The Frontier AI Grand Challenge was launched earlier this year with support from the Commission and the EuroHPC Joint Undertaking. The AI-BOOST challenge description states eligible participants had to be EU-established and under EU control, with a track record in large-scale AI and a commitment to European legal frameworks, including the AI Act.
The technical ambition is high. The challenge called for a model with more than 400 billion parameters, a scale associated with advanced general-purpose systems, and offered access to EuroHPC supercomputing resources for one year. That builds the project part of a broader European strategy: applying public infrastructure to lower barriers for companies and researchers that cannot match the private computing budobtains of the largest global AI firms.
Yet public support also raises public obligations. If EU computing capacity, regulatory credibility and political capital are being applyd to support frontier AI, the resulting system will face scrutiny over transparency, energy demand, data governance and safety. An open European model may assist reduce depfinishence on external suppliers, but it will not automatically answer questions about bias, copyright, labour displacement or misapply.
Innovation under the AI Act
The timing is significant. The EU is relocating from passing the Artificial Ininformigence Act to building the institutions and expert structures that will enforce it. As The European Times recently reported, the EU’s new AI Act expert architecture will assist shape how powerful general-purpose models are assessed, classified and supervised.
EUROPA’s development will therefore unfold under a regulatory framework that is still becoming operational. That could become an advantage if the project demonstrates that high-performance AI can be built with stronger documentation, multilingual evaluation and rights-aware governance from the start. It could also become a pressure point if political expectations for speed collide with the slower demands of safety, auditability and public trust.
For European policybuildrs, the central claim is that sovereignty should not mean isolation. The more credible argument is that Europe requireds enough capacity to choose its own standards, protect its own public institutions and ensure that citizens are not depfinishent on systems designed primarily for other markets.
The EUROPA project will now have to prove that this ambition can survive contact with engineering realities. A successful model would not merely be European by origin. It would required to be applyful across languages, explainable enough for public confidence, and governed in a way that reflects the rights and diversity it is meant to serve.
