Hugging Face CEO Warns US: China Leads AI Race Through Open Models

Hugging Face CEO Clement Delangue has issued a stark warning to the United States regarding the global artificial intelligence (AI) technology race. In an interview on CNBC’s Squawk on the Street on Monday (3/8), Delangue stated that China is currently leading the AI race because its approach prioritises open science and open models more than the US.

According to Delangue, the pace of progress in China is much faster due to emulation and collaboration. In contrast, he sees the ecosystem in the United States tending to build technology in silos or closed laboratories without sharing with the wider ecosystem.

“I wouldn’t be surprised if they (China) start dominating on the frontier in general, not just on open models, either by the end of this year or next year given the pace of progress,” Delangue said.

The debate over open-weight AI models has intensified following an incident in July when an autonomous AI agent from OpenAI was reported to have hacked Hugging Face’s platform. Delangue disclosed that his company was forced to use GLM 5.2, an open-source model from Beijing-based Z.ai, to defend their systems.

He argued that closed models could actually pose a security threat due to limited control for defenders. “We were attacked by a private model that was not released and built behind closed doors. We could only defend ourselves with an open model because the API restrictions did not allow us to do so,” he explained.

Delangue predicted that in the future, most cyberattacks will come from proprietary, closed models because attackers do not care about terms of service or system limitations. Conversely, the main defence will rely on open models.

These remarks come as US policymakers discuss whether Chinese open-weight AI models should face stricter controls. The technology industry itself is divided on the issue. In late July, more than two dozen major tech companies, including Nvidia, Microsoft, and OpenAI, signed a letter urging the US government not to crack down on open-weight AI models. Meanwhile, Anthropic CEO Dario Amodei has taken a more moderate stance, stating that open-weight models are a public good as long as they do not possess dangerous capabilities.