Clément Delangue, CEO of the AI platform Hugging Face, delivered a stark warning to investors and policymakers on Monday, stating that China is winning the artificial intelligence race precisely because its development ecosystem is more open than that of the United States.
“Everyone is building in silos in some of the frontier labs and not sharing with the rest of the ecosystem,” Delangue said during a CNBC appearance, contrasting the proprietary hoarding of American firms with China’s state-aligned push for open-weight models. He projected that China could begin dominating at the frontier of AI advancement by the end of the year or next, a timeline that signals a direct threat to American technological primacy and the economic leverage it provides domestic workers.
National Security Through Open Code
The debate arrives as the U.S. weighs export controls and restrictions on open-source AI, a move large American firms such as Meta and Nvidia have lobbied against. Delangue cited a recent security incident where a proprietary, unreleased model from OpenAI breached Hugging Face's defenses. He claimed the company could only repel the attack using open-source tooling, specifically citing GLM 5.2, a model developed by the Beijing-based firm Z.ai.
“We were attacked by an unreleased private model built behind closed doors,” Delangue stated. “And we could only defend ourselves with open models because the guardrails of the APIs didn't let us.” He argued that foreign adversaries will naturally ignore “terms of service” and corporate safety protocols, making proprietary models an offensive threat and open models the only viable defense mechanism for U.S. infrastructure.
The Cost of Isolationism
A broad industry coalition, including corporate giants like Microsoft, recently petitioned the government to avoid cracking down on open-weight AI. The current trajectory—where domestic frontier labs operate with extreme secrecy—risks paralyzing the broader American AI supply chain. As policymakers target Chinese tech with tariffs and export blacklists, the refusal of Silicon Valley’s major players to share advancements domestically could paradoxically accelerate the U.S. loss of strategic advantage, harming industries that employ millions of American workers dependent on a secure technological edge.
Delangue concluded that the future of AI conflict will see attacks originating from private, black-box models, with defenses mounted entirely by transparent, open systems. For the U.S., the path to maintaining hegemony may require breaking down the very information silos its largest lobbying entities have constructed.