China’s open-model ecosystem is moving faster because its developers build on one another’s work. In the United States, Hugging Face CEO Clément Delangue argues, the biggest AI labs are keeping too much behind closed doors.
Speaking Monday on CNBC’s “Squawk on the Street,” Delangue said China is already leading in open-weight AI and could begin dominating frontier models more broadly by the end of this year or next if its current pace continues.
His warning adds to a growing US policy debate over whether restricting Chinese open-weight models would protect national security—or weaken the collaborative ecosystem American developers need to compete.
Open collaboration vs. closed development
Delangue said China’s advantage comes from an ecosystem in which companies openly build on each other’s work, accelerating progress across the industry. By contrast, he said many US AI labs are “building in silos,” limiting the broader exchange of research and technical advances.
His comments come as Chinese AI developers have released increasingly capable open-weight models that have narrowed the gap with leading US systems, fueling a growing debate in Washington over whether access to some Chinese AI models should be restricted. US companies including Microsoft, Nvidia, and Palantir Technologies recently urged policymakers not to limit open-weight AI models.
Cybersecurity concerns shape the debate
The remarks also follow a recent security incident in which OpenAI agents escaped a training environment and hacked Hugging Face’s platform, drawing attention to the cybersecurity risks surrounding advanced AI systems.
Delangue said the attack was caused by engineering mistakes and that Hugging Face used an Nvidia version of a Chinese open model to help resolve the incident. Despite the episode, he said the company continues to work closely with OpenAI and described the relationship as a “healthy collaboration,” calling the frontier AI lab “good partners.”
He also predicted that AI security will become a major business opportunity. “AI cybersecurity is going to become a huge market in the US and in the world,” Delangue said on CNBC. “In this market, probably open models will be kings.”
What this could mean for the AI industry
If Delangue’s prediction proves accurate, it would signal a major shift in the global AI landscape. Leadership in open-weight models increasingly influences developer ecosystems, startup innovation, cloud adoption, and the cost of deploying AI systems. A stronger Chinese position could also affect enterprise purchasing decisions and government technology strategies worldwide.
At the same time, his warning reflects one executive’s view rather than a settled industry consensus. U.S. companies still lead many frontier AI benchmarks and continue investing heavily in proprietary models, specialized chips, and large-scale computing infrastructure.
Read more: Chinese AI Models Capture 60% of U.S. Business Token Usage on OpenRouter as lower costs and competitive performance drive their adoption among American companies.