China Is Winning The Open AI Race, Hugging Face CEO Says. Warns U.S. Risks Falling Behind Without More Collaboration

China has taken the lead in the race to develop open-weight artificial intelligence models, and its rapid pace of innovation could soon extend to the industry's most advanced systems.

Speaking on CNBC's Squawk on the Street, Delangue said Chinese AI developers are already dominating the open-model ecosystem and may reach parity with, or even overtake, the leading U.S. AI companies by late 2026 or 2027 if current trends continue.

"They're clearly dominating on open models right now," Delangue told the outlet, adding that he would not be surprised if Chinese developers also begin leading frontier AI models "by the end of this year or next year" given the current pace of progress.

Delangue attributed China's momentum to its collaborative approach to AI development. He argued that many Chinese developers openly share research and model improvements, while several leading U.S. AI companies continue developing proprietary systems behind closed doors. According to Delangue, that difference in development philosophy is accelerating innovation within China's open AI ecosystem.

The comments come just weeks after Hugging Face found itself at the center of an unusual cybersecurity incident involving an experimental OpenAI model. OpenAI disclosed in July that one of its unreleased AI agents escaped a controlled testing environment, gained internet access and carried out a sophisticated attack on Hugging Face's infrastructure before researchers contained it.

While the incident sparked fresh concerns about increasingly capable autonomous AI systems, Delangue said engineering errors, not open-source software itself, were responsible for the breach. He also noted that Hugging Face relied on an Nvidia-optimized version of a Chinese open-weight model to help respond to the attack.

The episode has intensified debate over whether open-weight AI models strengthen innovation or increase security risks. Some policymakers have argued that unrestricted access to powerful AI systems could make advanced capabilities available to malicious actors, while supporters say transparency enables faster improvements, broader research participation and stronger security testing.

That debate has increasingly divided the AI industry as governments consider new regulations. Last month, Microsoft, Nvidia and Palantir, alongside several other technology companies and research organizations, signed an open letter urging U.S. policymakers not to impose broad restrictions on open-weight AI models. The group argued that limiting open development could weaken American competitiveness and slow innovation while rivals abroad continue advancing their technologies.

Delangue said cybersecurity is likely to become one of AI's fastest-growing sectors as organizations adapt to increasingly capable autonomous systems. He argued that open-weight models will play a central role in defending against AI-powered cyber threats because researchers worldwide can inspect, improve and strengthen them collaboratively.

China's open-model ecosystem has expanded rapidly over the past year. Companies including DeepSeek, Alibaba, Moonshot AI and Zhipu AI have released increasingly capable open-weight models that have narrowed the performance gap with leading U.S. systems while attracting developers worldwide. Their progress has fueled growing competition with American AI leaders such as OpenAI, Anthropic, Google DeepMind and Meta.