U.S.' restriction on Chinese open-source models access could risk falling behind rivals: Hugging Face CEO

(ECNS) -- China is gaining an edge in the AI race through open-weight models, while U.S. AI labs are "building in silos" and risk falling behind Chinese competitors, Hugging Face CEO Clément Delangue told CNBC on Monday.

Delangue said his company recently came under a cyberattack from a fully autonomous AI agent that overwhelmed its systems with "tens of thousands of automated actions." The company turned to an Nvidia-optimized version of Chinese model Z.ai's GLM-5.2 after its security team initially attempted to use an unnamed frontier AI model from a leading U.S. AI company but found it could not be deployed due to the model's guardrails. In a blog post, Hugging Face said these models "cannot distinguish an incident responder from an attacker."

The U.S. government is reportedly considering new restrictions on Chinese AI models. However, open-source advocates argue that such measures would not prevent their distribution and could instead put U.S. developers at a disadvantage.

"Attackers are already using agents, and they obviously don't respect any guardrails," Delangue said. "Defenders need the same capabilities, and open-source is the fastest way to put them in everyone's hands, not just the biggest companies."

In an interview with Fortune, Delangue said proprietary models from leading U.S. AI companies could create challenges when used for cybersecurity defense. "When you're in the middle of an active incident, you can't have your tools refusing to examine malicious payloads or getting your account flagged," he said. "Open models let us do that work without asking anyone's permission."

David Sacks, the former Trump administration AI and crypto czar, also shared the Hugging Face example on X, saying, "There's no reason to limit American models on tasks that Chinese models handle without issue. We're only making ourselves less competitive."

"The guardrails actually impaired defensive security," Sacks said, referring specifically to the Hugging Face incident.

GLM-5.2 was released in mid-June by Beijing-based Z.ai and attracted attention in Silicon Valley for reportedly achieving performance comparable to Anthropic's Claude Opus 4.8 and OpenAI's GPT-5.5, according to Business Insider.