The European Union Cybersecurity Agency and its research center (JRC) have begun testing open AI models from China. According to EU officials, this initiative is part of a broader research project to evaluate artificial intelligence, which also includes advanced models from the United States and Europe.
Purpose of the Tests
The tests conducted on Chinese models at the JRC include a comparative scientific assessment of the capabilities, performance, limitations, and risks of these models. Additionally, the JRC has examined the potential scientific applications of these models. However, the main focus of the JRC tests is not on the safety and security of artificial intelligence but rather on the scientific comparison of different models.
Read more: Competition for Key EU Diplomatic Positions Begins
Open Weight Models and Technology Competition
Open weight models are those whose main components are publicly released, allowing anyone to run and study them. China's approach to artificial intelligence leans more towards open models, while some major technology and AI companies in the United States, including NVIDIA, are investing in this type of model.
Currently, the United States and China are in a competition for dominance in new technologies, with Europe largely on the sidelines. Without a viable open model from Europe, European private and public organizations may have to choose between American and Chinese suppliers for advanced AI applications.
In a recent speech, Ursula von der Leyen, President of the European Commission, suggested that Europe should not try to win the race for frontier AI. Additionally, a major German technology company has recently announced that instead of collaborating with the leading European AI lab, Mistral, it has turned to Chinese open models.
The JRC has not engaged in any negotiations regarding specific access or licensing with Chinese developers, and the models are run on the secure and internal scientific infrastructure of this center. Furthermore, the European Commission does not intend to use Chinese models for its purposes.
The European Union Cyber Agency (ENISA), which has gained access to advanced models from major U.S. AI companies, is also testing several open weight Chinese models. These two entities, JRC and ENISA, are required to collaborate to create an AI testing framework under the recent EU action plan on AI and cyber.
Read more: Europe Prepares for Deadly Russian Attacks
