Anthropic goes to made every effort to prevent Chinese citizens from using AI models, but in practice the protections often failed. Over the past year, startups, researchers and technology enthusiasts across the country have developed increasingly sophisticated ways to access Claude. Many of them consider it the most powerful AI assistant in the world, so it’s worth the extra effort to get it.
In early June, Anthropic publicly released Fable 5, a secured version of its most powerful AI model to date, Mythos. Chinese social media immediately filled with posts from people who shared their impressions after trying this solution. (A few days later, Anthropic revoked access to the model globally in response to export controls imposed by the Trump administration.)
Chinese generally have access to other Western AI tools such as OpenAI’s ChatGPT, using virtual private networks, foreign phone numbers and international payment methods to create and maintain their accounts. However, Anthropic has likely taken more aggressive steps, such as blocking accounts it believes are owned and controlled by people located in China. On Chinese social media, users often report that they have been suspended from Claude without warning, despite taking these precautions.
A game of cat and mouse gave Claude access to China’s growing underground economy. The accounts are sold on Chinese e-commerce platforms such as Taobao and on illegal marketplaces on Telegram. A cottage industry of “transfer stations” has also recently emerged. These services act as intermediaries, purchasing access to the Anthropic API outside China and then redistributing Claude API tokens to users in that country. The setup is intended to provide startups and other professional users with more stable and reliable access to the AI assistant.
Michael Aciman, a spokesman for Anthropic, says the company uses a number of evolving detection systems, including identity verification, to enforce its policies to prevent unauthorized access to Claude. He added that Anthropic also worked to detect and disrupt proxy networks used to provide access to the chatbot in China.
Despite all the difficulties that the Chinese have to overcome to utilize Claude, there remain many steadfast Anthropic fans in the country. It is especially popular among developers. Even though Chinese companies like DeepSeek and Z.ai have some of the most powerful open-source, large-language models on the market, third-party testing still shows that they lag behind leading closed-source models like Claude. During a recent reporting trip to China, WIRED spoke with scientists and engineers from a number of technology companies who said they preferred using Claude to Chinese models for code generation and were eager to try out any fresh model Anthropic releases.
Zilan Qian, a research fellow at the Oxford China Policy Lab, looked into the situation black market for reselling Western AI tokens to Chinese users. He noted that Chinese software developers say they overwhelmingly prefer to utilize tools like Claude Code and Codex OpenAI over those from domestic companies. “The analysis shows that Chinese models still lag six to nine months behind US models, and when it comes to specific issues such as coding and programming, the difference is clearly visible,” says Qian.
“Both Chinese AI policymakers and technology professionals have much less difficulty drawing on and using American ideas or products, regardless of geopolitical or ideological rivalries,” says Matt Sheehan, a senior research fellow at the Carnegie Endowment for International Peace, where he studies AI policy and China. “It’s Americans who tend to think that an idea or product is tainted just because it comes from their rival,” he says.
Dario Amodei, co-founder and CEO of Anthropic, often clearly considers China’s access to frontier models a critical threat to US national security. This week only, Anthropic accused Alibaba using Claude’s results to train the Chinese company’s competing models, a technique known as “distillation.” Anthropic also claimed other Chinese companies have done the same in the past. For this and other national security reasons, Anthropic does not offer commercial access to Claude in China or to subsidiaries Chinese companies located outside the country.