EU research centre and cyber agency test Chinese AI models
The Joint Research Centre is comparing Chinese models with US and European alternatives, while ENISA is also experimenting with open-weight systems.
Two EU bodies have begun testing Chinese artificial intelligence models. European Commission spokesperson Maciej Berestecki confirmed the work at the Joint Research Centre (JRC), while Juhan Lepassaar, executive director of the EU cybersecurity agency ENISA, confirmed similar experiments at his agency.
Berestecki said the JRC’s research covers publicly accessible open-weight systems from China alongside leading European and US models. Researchers are comparing what the models can do, how they perform, their shortcomings and risks, as well as possible uses in scientific work.
The JRC runs the models on its own secure scientific infrastructure, Berestecki said. It has not negotiated special access or licences with the Chinese developers, and the Commission does not intend to adopt Chinese models for its own use.
A separate Commission official, speaking anonymously because of the sensitivity of the material, said AI safety and security were not the main priorities of the JRC experiments. Open-weight models make their core components public, enabling others to operate and examine them.
Lepassaar said ENISA was trying a small number of Chinese open-weight models. The agency also has access to advanced systems developed by Anthropic, OpenAI and Google. ENISA and the JRC have been assigned to jointly establish an AI testing platform as part of a recent EU action plan covering AI and cybersecurity.
