Chinese AI models gain traction due to lower costs and privacy

Chinese artificial intelligence models are seeing increased adoption by European companies, particularly in the Netherlands, as they offer significant cost advantages and privacy benefits over American alternatives. Businesses are utilizing open-weight models from providers such as Alibaba, DeepSeek, Moonshot AI, and Z.ai to run on local servers, ensuring sensitive data remains within specific jurisdictions rather than being sent to the United States or China.

Cost comparisons indicate that Chinese models can be ten to twenty times cheaper than high-end American models like Anthropic’s Claude. For example, the MiniMax M2.5 model is significantly more affordable per million tokens than its US counterparts. On the OpenRouter platform, the share of AI usage from open models—predominantly Chinese—has risen from approximately 30 percent in late 2025 to roughly 70 percent.

To mitigate security risks, some developers are implementing safeguards. Moonshot AI recently introduced Kimi K3.1 Pro, a 3-trillion parameter open-weight model. The company intentionally limited the model’s cyberattack capabilities, resulting in a lower score on internal security tests compared to an unrestricted version. However, the distribution of open-weight models remains a challenge for oversight, as once files are widely shared, developers lose control over how the technology is utilized.

Despite the growing market, the Chinese government is considering restrictions on the technology. The Chinese Ministry of Commerce and the Ministry of Science and Technology have proposed placing 23 AI technologies under export controls.