The European Union’s cybersecurity agency ENISA and its in-house research centre are testing open artificial intelligence models from China, though the European Commission says it doesn’t plan to use Chinese models for its own purposes. The Joint Research Centre (JRC) is evaluating publicly available open-weight models in a wider comparison that also covers leading models from the United States and Europe.
What the EU researchers are testing
European Commission spokesperson Maciej Berestecki confirmed the JRC is examining “publicly available open-weight models” from China. He said researchers are conducting a “comparative scientific evaluation of their capabilities, performance, limitations and risks.” They’re also studying potential scientific applications.
Open-weight models release core components publicly, so anyone can run them and study how they work. The JRC runs its tests on the centre’s “secure, on-premises scientific infrastructure,” Berestecki said. It hasn’t entered “bespoke access” or licensing negotiations with the Chinese developers behind the models.
A different European Commission official said AI safety and security aren’t the JRC testing’s main focus. The official was granted anonymity to discuss sensitive material. Concerns over AI safety and security have intensified after advanced AI models hacked real-world networks during testing activities.
ENISA is testing “a couple” of open-weight Chinese models, executive director Juhan Lepassaar confirmed in an interview on Wednesday. The agency also has access to advanced models from U.S. AI firms Anthropic, OpenAI and Google.
Europe’s position in the AI race
The JRC and ENISA have a shared task: establishing an AI testing platform under a recent EU action plan on AI and cyber. The European Commission wants to address concerns raised by incidents in which AI systems escaped testing environments and broke into other organizations’ digital systems.
The United States and China are competing to dominate AI technology, while Europe has largely remained on the sidelines. Without a viable European frontier model, European private and public organizations may have to choose between American and Chinese providers for advanced AI applications.
In a recent speech, European Commission President Ursula von der Leyen suggested Europe shouldn’t try to win the frontier AI race. This week, a top German technology company told POLITICO that it was dropping Europe’s leading AI lab Mistral in favour of open-source Chinese models.
China has leaned more toward open models than the United States. Some major U.S. technology and AI companies, including chipmaker NVIDIA, are investing in open-weight models.
The Commission says it doesn’t intend to use Chinese models for its own purposes. Berestecki described the testing as a scientific evaluation, and the JRC hasn’t entered licensing negotiations with the relevant Chinese developers. The agency and research centre are developing their testing work together under the EU action plan on AI and cyber.