Claude's Great Wall Problem: How China's Underground AI Economy Keeps Beating Anthropic's Blocks

Despite aggressive geolocation restrictions and account bans, Chinese developers, researchers, and startups have built a thriving underground economy to access Anthropic's Claude. From Taobao account resellers to API 'transfer stations,' the cat-and-mouse game shows no signs of slowing down.

Anthropic has gone further than almost any other Western AI company to block users in China from accessing its models — yet a sophisticated, layered ecosystem of workarounds has emerged to defeat those controls at every turn.

Why Chinese Users Want Claude So Badly

Chinese AI models from companies like DeepSeek and Z.ai are among the most capable open-source options available — but they still trail the leading closed models. According to Zilan Qian, a research associate at the Oxford China Policy Lab, Chinese developers overwhelmingly prefer tools like Claude Code and OpenAI's Codex for programming tasks.

"Analysis shows that Chinese models are still six to nine months behind the US models, and for specific things like coding and developing, you can obviously tell the gap," Qian says.

During a recent reporting trip to China, WIRED spoke with academics and engineers at multiple tech firms who confirmed they regularly use Claude to generate code — and eagerly test each new model release.

The Restrictions Anthropic Has Put in Place

Unlike OpenAI, which Chinese users can access with a VPN, foreign phone number, and international payment method, Anthropic has taken more aggressive enforcement steps:

Account suspensions targeting users suspected of connecting from China
Identity verification via third-party service Persona (backed by Founders Fund), requiring government-issued photo ID
Ongoing detection and disruption of proxy networks
No commercial access for Chinese companies or their foreign subsidiaries

Michael Aciman, an Anthropic spokesperson, confirmed the company uses "a range of evolving detection systems, including identity verification" to enforce its policies.

The Underground Economy That Emerged Anyway

Restrictions have spawned a thriving black market. The main channels include:

Taobao and Xianyu — Chinese ecommerce platforms where pre-configured Claude accounts are sold outright
Telegram marketplaces — Underground channels selling Claude Pro, Claude Max, ChatGPT Plus, and Gemini Plus accounts, according to cybercrime investigator Hieu Minh Ngo of Vietnamese nonprofit ChongLuaDao
Transfer stations (relay stations) — Middlemen who purchase API access from Anthropic-approved countries, then resell tokens to Chinese users at discounted rates, often undercutting Anthropic's own pricing via enterprise deals

Transfer stations route requests through servers in Anthropic-supported countries. To end users, the experience is nearly indistinguishable from using Claude directly — just on a different platform.

The model has become popular enough that Chinese-language websites and GitHub pages now list dozens of competing transfer stations, comparing models and token prices. Even crypto billionaire Justin Sun launched his own transfer station in May 2025.

A Distorted Global Picture

The volume of Chinese proxy traffic may be skewing Anthropic's own usage data. Singapore — a country of just 6 million people with strong Chinese-language ties and international business prominence — frequently ranks among the top countries globally for Claude adoption relative to population.

The explanation is straightforward: Chinese users disproportionately route traffic through Singapore to simulate a consistent, plausible location.

The Geopolitical Stakes

Dario Amodei, Anthropic's cofounder and CEO, has repeatedly identified Chinese access to frontier AI as a national security threat. This week, Anthropic accused Alibaba of using Claude outputs to train its own competing models via a technique called distillation — and alleged other Chinese companies have done the same.

Not everyone sees the situation through an adversarial lens, however.

"For both Chinese AI policymakers and technical people, they have much less of a problem drawing on and using American ideas or products, regardless of the geopolitical or ideological rivalry. It's Americans who tend to think an idea or a product is tainted just because it comes from their rival," says Matt Sheehan, senior fellow at the Carnegie Endowment for International Peace.

The restrictions are tightening — but as long as Claude remains the preferred tool for serious developers, the demand driving this underground ecosystem isn't going anywhere.