According to a People's Daily report on August 10, an industry association comprising nearly 200 Silicon Valley tech startups sent a joint letter to the US government, explicitly opposing restrictions on domestic companies' use of Chinese open-weight AI models. The letter stated plainly: "US developers need continued access to the full range of open models available globally."
Chinese Open-Source Models Shift from "Option" to "Must-Have"
The report cited third-party industry monitoring data indicating that US companies' token usage of Chinese AI models has consistently remained above 30% weekly, peaking at 46%. From medical startups accelerating technological iteration to digital service platforms using them as default development tools, Chinese open-source models are providing practical solutions for industrial innovation across various countries. Bloomberg analysis suggested that comprehensive restrictions on Chinese open-source models entering the US market would significantly impact related businesses and users.
The joint letter also warned: "Restrictive measures would stifle competition, entrench incumbents, and effectively impose a tax on intelligence—raising costs and narrowing choices."
US "Wall-Building" Tactics Escalate, Companies Unite in Rare Opposition
CCTV's "Blue Hall Observation" program reviewed the US's intensified actions over the past two months: first, nearly ten Chinese AI companies, universities, and research institutions were added to restrictive lists; then, the Federal Communications Commission took measures against humanoid robots and quadruped intelligent robots. The lists have expanded continuously—from talent and companies to large models, compute infrastructure, and application ecosystems.
Professor Shen Yi from Fudan University's Cyberspace Governance Research Base analyzed that US politicians' logic is "the US must monopolize"—by monopolizing AI and compute, they can once again monopolize knowledge, wealth, and power. However, China not only provides cost-effective alternatives through open-source models but also promotes global AI governance, treating AI as a public good for humanity—something US politicians find "unacceptable."
US companies' rare collective "endorsement" of open source this week is noteworthy. In July, after senior US officials attempted to label Chinese open-source models with the "distillation" concept, more than 20 US tech companies including NVIDIA, Microsoft, and Meta co-signed an open letter explicitly opposing a "one-size-fits-all" crackdown. NVIDIA CEO Jensen Huang stated directly: "About half of the world's AI researchers are in China. China's models are excellent. Open-source models are great and should be used globally."
The 200-startup joint letter and Hugging Face's use of China's GLM-5.2 for "rescue" form a closed loop—US companies are "voting with their feet" for Chinese open-source models. While the US government builds walls in Washington, Silicon Valley engineers are writing code with Chinese models. Sanctions not only fail to contain China's AI momentum but also cut off domestic startups' technology supply and raise innovation costs. As the joint letter states, restrictions "effectively impose a tax on intelligence."