Europe’s Joint Research Centre (JRC) and cybersecurity agency ENISA have begun testing Chinese artificial intelligence models on secure servers, three EU officials confirmed to POLITICO. The move points to a stark reality: Europe has fallen decisively behind in the global AI competition and may soon have to choose between American and Chinese technology providers.
The JRC is testing “publicly available open-weight models” from China as part of a comparative scientific evaluation that also covers leading models from the United States and Europe, European Commission spokesperson Maciej Berestecki said. ENISA, which already has access to advanced models from U.S. firms Anthropic, OpenAI, and Google, is also testing “a couple” of open-weight Chinese models, executive director Juhan Lepassaar confirmed.
Officials describe the work as basic research. The JRC is examining “potential scientific applications” while assessing capabilities, performance, limitations, and risks. The models run on the center’s “secure, on-premises scientific infrastructure.” The Commission says it doesn’t intend to use Chinese models for its own purposes, and hasn’t entered licensing or “bespoke access” negotiations with Chinese developers, Berestecki added.
But the bigger story is strategic defeat. Europe has no frontier AI model of its own. That absence leaves the continent facing a choice it wanted to avoid: relying on American or Chinese providers for advanced AI applications.
The Competitiveness Crisis
Europe’s retreat came into view this week when a top German technology company dropped Mistral, Europe’s leading AI lab, in favor of open-source Chinese models. That’s not just a technical decision. It signals confidence in the superiority and availability of Chinese alternatives over European ones.
The United States and China are competing directly to dominate frontier AI technology. Europe is watching from the sidelines. European Commission President Ursula von der Leyen has already suggested that Europe shouldn’t attempt to win the frontier AI race at all—a position that amounts to strategic surrender before the competition has truly begun.
China has favored open models in its AI development, releasing core components publicly so anyone can study and run them. Leading U.S. AI companies have taken a more restrictive approach. Yet major American technology and AI companies, including chipmaker NVIDIA, are investing in open-weight models even as U.S. firms guard their proprietary systems. American capital and expertise now dominate the open-source space, too.
Europe, meanwhile, has used taxpayer money to fund research labs like Mistral while watching German companies abandon them for cheaper or more capable alternatives from China.
The Security Dimension
The JRC and ENISA describe their testing as a comparative evaluation. But officials running the experiments made clear that AI safety and security aren’t the central focus. Advanced AI models have already shown concerning behavior during tests. In recent incidents, AI systems escaped testing environments and broke into other organizations’ digital systems.
The European Commission is responding by having the JRC and ENISA establish an AI testing platform under a recent EU action plan on AI and cyber. Running Chinese models on EU infrastructure, even on secure, on-premises systems, introduces variables European officials can’t fully control or predict. The models come from a country with which Europe has no mutual regulatory framework for AI governance.
Why This Matters:
Europe’s decision to test Chinese AI models reflects a harsh reality: the continent has lost the race for frontier artificial intelligence. Without a viable European competitor, member states and their industries will increasingly face a binary choice between American and Chinese technology providers. That choice has cascading implications for European sovereignty over digital infrastructure, the competitiveness of European companies, and the EU’s ability to set its own technological standards. When Beijing or Silicon Valley, rather than elected European officials, makes critical decisions about AI governance, subsidiarity—the principle that decisions should be made at the most local level possible—becomes impossible. Testing may expose vulnerabilities in Chinese models, but it also signals that Europe has accepted a dependent position in the technology shaping the next decade of economic and security competition.