AI Haven
AI News

US Government Advisory Body Warns China's Open-Source AI Models Now Outpace US

USCC warns China's open-source AI models now lead global usage, surpassing Meta's Llama in downloads and powering 80% of US AI startups.

March 24, 2026

The U.S.-China Economic and Security Review Commission (USCC) issued a stark warning on March 23, 2026: China's dominance in open-source AI models now threatens America's lead in artificial intelligence, driven by a self-reinforcing competitive advantage that export controls have failed to counter.

Open-Source Models Surpass US Counterparts

Chinese large language models from Alibaba, DeepSeek, Zhipu AI, and Moonshot AI now lead worldwide usage rankings on platforms like Hugging Face and OpenRouter, surpassing U.S. models in both monthly and cumulative downloads. Alibaba's Qwen series has eclipsed Meta's Llama in total downloads, while DeepSeek's R1 became the top U.S. App Store download following its release.

"The open-source approach enables China's cost advantages and broad global adoption, challenging U.S. closed-system dominance," the commission noted. The shift represents a fundamental change in how AI technology spreads worldwide.

80% of US AI Startups Rely on Chinese Models

Approximately 80% of U.S. AI startups now use Chinese open-source models due to their cost efficiency and performance approaching Western counterparts, despite existing U.S. export controls on advanced chips. This adoption rate underscores how the software layer has decoupled from hardware restrictions.

The competitive landscape has shifted dramatically: Qwen models exceeded 750 million downloads by late 2025, outpacing all competitors combined in monthly metrics. DeepSeek's reasoning models (R1 and distilled variants) dominate adoption on platforms like LM Studio and Featherless.ai, with some variants exceeding one million downloads.

The Deployment Feedback Loop

Beyond model performance, the commission highlighted a compounding advantage in embodied AI. China's integration of AI into factories, logistics, and robotics generates real-world data that further refines models—a "deployment gap" that favors China in robotics and autonomous systems.

Beijing has prioritized embodied AI as a strategic industry, with leading robotics firms planning public listings in 2026. This stands in contrast to the U.S. focus on proprietary models, which limits deployment data collection.

The implications extend beyond AI capability: China's open-source strategy enables global infrastructure building through cloud platforms like Alibaba, Huawei, and Tencent, while the U.S. relies on API access that remains less accessible to developers worldwide.

Source: ReutersView original →