How People in China Keep Outsmarting Anthropic’s Geolocation Restrictions

A Chinese software developer, sipping tea in a quiet café in Hangzhou, logs into an AI chatbot through a local proxy server. The screen flickers once, then settles into a familiar interface — the one that should not be accessible from within China. This is just one of many daily acts of digital subterfuge, as users navigate the invisible walls built by Anthropic to keep its advanced AI models out of the country.

The Underground Economy of AI Access

In response to Anthropic's geolocation restrictions, a growing number of users in China have turned to an underground economy to access its AI chatbot, Claude. Proxy services, forged identities, and intermediary "transfer stations" have become common tools in this digital arms race. On platforms like Taobao and Telegram, accounts with verified access to Claude are sold in batches, often accompanied by step-by-step guides to bypass detection systems.

  • Users buy pre-configured Claude accounts that have already passed identity verification checks.
  • Transfer stations offer a more stable solution for developers, routing prompts and responses through API keys located in supported regions.
  • Some sellers even provide fake government-issued IDs to help users bypass Anthropic’s latest verification requirements.

This ecosystem has grown rapidly, fueled by the demand for tools that outperform domestic alternatives. Chinese programmers, in particular, favor Claude Code for its superior ability in generating complex code, a task where local models still fall short.

The Cat-and-Mouse Game with AI Restrictions

Anthropic has implemented a range of countermeasures to deter unauthorized access, including identity verification systems and detection of proxy networks. A spokesperson for the company confirmed these efforts, stating that Anthropic is constantly updating its policies to prevent abuse. Yet, for every new layer of security, users in China have found a workaround.

The rise of "relay stations" has been one of the most effective strategies. These services operate from outside China, allowing users to interact with Claude as if they were logged in directly. Unlike traditional proxies, which can be easily flagged, these intermediaries offer a smoother experience, mimicking the behavior of legitimate users and reducing the risk of detection.

In some cases, users have even begun to route traffic through Singapore, where the number of Claude users is disproportionately high. This has led to a distorted global usage map, as the true number of Chinese users remains hidden behind layers of obfuscation and intermediaries.

The Future of AI Access in China

As Anthropic's restrictions continue to evolve, so too will the methods used to circumvent them. Experts warn that the proliferation of these unauthorized access tools poses new risks. Users may be exposed to scams, data theft, or even the misuse of AI-generated content. Researchers like Zilan Qian emphasize that the black market for AI access is not only a technical challenge but also a growing concern for AI safety and data privacy.

Despite the risks, the demand for advanced AI tools in China shows no sign of slowing. With OpenClaw and other AI agents consuming vast amounts of tokens, the need for reliable and affordable access has become more urgent than ever. As long as Anthropic and its competitors continue to limit access based on geography, users in restricted regions will find ways to bypass the barriers — and the underground economy will continue to thrive.

The battle over AI access is not just a technical one. It reflects a broader struggle between innovation and control, between users seeking the best tools and companies trying to enforce geographic boundaries. In this ongoing game, the people in China are not just outsmarting the rules — they are redefining how the world accesses the future of artificial intelligence.