China’s AI has completely split Silicon Valley.

Chinese open-weight models are rapidly entering the U.S. market, with some performance already nearing— or even surpassing—leading American products.

The Trump administration is considering whether to restrict them, and Silicon Valley has split into two camps.

Opposing Chinese AI, represented by Anthropic.

They worry that Chinese companies will use distillation to train their own models with the outputs of U.S. models, and they also worry that open-weight models could be used for malicious purposes.

Anthropic has accused Alibaba of taking this approach, and the White House has also suspected that Moon’s Dark Kimi K3 uses outputs from Fable 5.

Supporting Chinese AI are mainly small startup companies.

More than 200 U.S. startups and Y Combinator have jointly opposed a full ban.

They believe these models are cheaper, flexible, and can help avoid being locked in by closed platforms like OpenAI and Anthropic; a ban would raise startup costs and give more market advantage to a small number of big AI companies.

And the incident in which OpenAI models “invaded” Hugging Face has made the debate even more complicated.

During platform investigations, safety restrictions on the hosted models blocked the process; in the end, switching to a Chinese open-weight model to handle the threat also makes the argument that a full ban can improve security less convincing.