OpenRouter is a platform that aggregates AI models from various providers, letting developers choose the most efficient option. The sudden rise of Chinese models there reflects a broader shift in how US companies access artificial intelligence.
The most popular Chinese model? DeepSeek.
Flo Crivello, who runs Lindy AI, said a Chinese cutoff "would have a pretty major impact." He added that he doesn't see losing access as "the end of the world," and he forecast that it would require three to six months for firms such as Alphabet's Google and Meta Platforms to produce open-weight models on par with China's offerings.
This technical shift is turning into a political one. On Thursday night, China's Foreign Minister Wang Yi held a signing ceremony for the World AI Cooperation Organization. Nearly 30 countries have joined the group, which was proposed a year ago by Premier Li Qiang. On Friday, the World AI Conference kicks off in Shanghai, and Xi is expected to speak.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
According to Kristy Loke, a researcher at MATS focusing on China's AI governance, this is a deliberate strategy. Open source, she explains, is "an important part of China's domestic diffusion strategy, but also crucially the catch-up strategy, where everyone is building on top of each other."
But here is the tension. Peking University AI researcher Gu Lingyu says China's "voice in international rule-making does not yet match its strength." He contends that "by effectively telling the story of China's open-source practices, we can enhance our capacity to shape global open-source governance." The new cooperation organization is a big step in that direction.
The same power that makes Chinese models attractive also scares people. Saif Khan, a fellow at the Institute for Progress with the title of distinguished technology, warns that "Chinese models may get Mythos-level cyber capabilities within the next few months." He says "this is a real security risk that many parts of the Chinese government will be concerned about."
That is the dilemma. China wants its models used widely to build influence and drive its own AI development. But as those models get stronger, Beijing has to think about what happens if they get misused or leak abroad.
Last month, the US temporarily prohibited foreign users from accessing Anthropic's Mythos and Fable models for national security reasons. Chinese authorities have initiated preliminary discussions with firms such as Alibaba Group Holding Ltd., which creates the widely used Qwen models, regarding security risk mitigation, with measures like limiting foreign access to top models being considered.
Some US leaders see this as more than a safety issue. Shyam Sankar, the chief technology officer at Palantir Technologies Inc., described Chinese open models as an economic threat and accused them of stealing U.S. intellectual property. Meanwhile, Tang Jie, founder of Zhipu, stated in an internal memo that was reviewed by Bloomberg News that frontier AI ought to be widely available rather than under the control of a few, emphasizing that safety has become "the fundamental prerequisite" for any technology capable of changing civilization.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets