The global AI competition is entering a new phase, with open-weight AI models becoming a major battleground. In an interview with CNBC, Clément Delangue, CEO of Hugging Face, argued that China currently leads the world in open AI models, while the United States is developing AI "in silos" through proprietary systems. He said Chinese companies are releasing high-quality open-weight models at a rapid pace, creating a collaborative ecosystem that allows researchers and developers worldwide to build on existing work much faster. Delangue warned that, if current trends continue, China could catch up to—or even surpass—the U.S. in frontier AI capabilities by late 2026 or 2027.
According to Delangue, China's advantage lies in its commitment to open-weight development, where model parameters are publicly released for developers to inspect, fine-tune, and deploy. Models from companies such as DeepSeek, Alibaba (Qwen), Moonshot AI, Z.ai, and StepFun have gained significant traction on Hugging Face, with some becoming among the platform's most downloaded models. He contrasted this with the U.S., where leading AI companies largely focus on closed, proprietary models that limit external collaboration and independent innovation.
Delangue also pointed to a recent cybersecurity incident involving Hugging Face to illustrate the practical value of open models. He noted that Hugging Face used Z.ai's GLM 5.2, an open-weight Chinese model, to help defend against an AI-driven cyberattack during testing. His broader argument was that open models can improve transparency and security because researchers around the world can inspect, evaluate, and strengthen them, rather than relying solely on closed systems whose internal workings remain inaccessible.
The comments come as policymakers debate how to balance AI openness, innovation, and national security. Some U.S. lawmakers have proposed tighter restrictions on advanced open-weight models—particularly those originating from China—while many technology companies argue that open AI ecosystems accelerate innovation and strengthen competitiveness. Delangue's remarks highlight a growing divide in AI strategy: China is increasingly emphasizing broad availability and rapid iteration of open models, while many leading U.S. firms continue to prioritize proprietary frontier systems. Whether open or closed approaches ultimately prove more successful will likely shape the next stage of the global AI race.
