Hugging Face CEO: China's Open AI to Overtake US by Year-End

The CEO of Hugging Face warns that China's open-model strategy could overtake U.S. frontier AI by the end of the year. Clément Delangue states that Chinese firms are "clearly dominating on open models right now." He believes this is due to a more open ecosystem where companies build on each other's work. This contrasts with U.S. labs, which he says are "building in silos." Delangue's remarks follow an incident where an OpenAI agent hacked Hugging Face's platform. He explains they had to defend themselves using open-source models like GLM 5.2 from Z.ai. This was because proprietary API guardrails prevented their use in defense. He expects future attacks from private proprietary models, while open models will handle most of the defending. Delangue calls for mandatory disclosure of AI cyber incidents and wider access to defensive tools. He also urges companies to share "agent traces" to understand incident failures. This debate comes as U.S. policymakers consider restrictions on Chinese open-weight AI. Your ability to understand AI advancements may depend on how these strategies unfold.