The Rise of Claude Tokens in China’s Underground Market

Claude has become a highly sought-after model among developers and engineers, yet it remains unavailable in mainland China due to regulatory and safety reasons. This unavailability has sparked the creation of a black market for Claude tokens, allowing local developers to access models like Claude Opus 4.7 while circumventing imposed restrictions.

Challenges in Payment Methods

No Paying with Alipay. One crucial measure Anthropic has implemented to restrict access is the exclusive acceptance of international credit cards like Visa and Mastercard. This policy excludes local payment methods such as Alipay and WeChat Pay, creating a significant hurdle for Chinese users. However, resourceful individuals have found ways to overcome this barrier.

Using Virtual Credit Cards

Virtual Cards. To bypass payment restrictions, many Chinese users are turning to virtual credit card (VCC) services, including DuPay and WildCard. These platforms facilitate the acquisition of Hong Kong or U.S. credit cards funded with cryptocurrencies or local transfers, thus allowing users to deceive Anthropic’s billing systems. Furthermore, they have also found innovative solutions for challenges like SMS verification, employing “SMS farms” to meet identity verification requirements imposed by Anthropic and others.

Connection Issues and Transfer Stations

The “Transfer Stations” (中转站). Even after overcoming payment hurdles, users face latency issues and connection disruptions that hinder Claude’s usability in China. To resolve this, “Transfer Stations” have emerged—servers acting as intermediaries between foreign servers and local users. These gateways relay requests from China to Anthropic’s servers, simulating requests from authorized locations. As a result, latency remains low, providing Chinese users with an experience comparable to users in the U.S. or Europe. These stations have gained notoriety and can be found listed on various platforms, including GitHub and ranking websites.

Excessive Discounting of Claude Tokens

Claude is Almost Free in China. Remarkably, these methods not only allow access to Claude but also drastically reduce its price, reaching as low as 5% or 10% of the original cost due to the aforementioned transfer stations. Developers can utilize Claude at an astonishing rate of one yuan for every dollar of tokens.

The Almond Tree Trick

This steep discount arises from several tactics, including:

  • Mass purchasing of server capacity.
  • Utilization of accounts created with stolen or fraudulent cards.
  • Exploiting promotional credits.
  • A crucial hook: providers may lose money with Claude tokens, but they attract developers by offering them more profitable local models like DeepSek.

The Risks of the Black Market

Am I Really Using Claude? One danger in this cheap token market is fraud. Some Chinese resellers falsely offer what they call the “Claude API,” providing subpar models instead. Users’ difficulty in identifying this deception increases unless they engage in complex tasks or have prior experience with the model.

Goodbye to Privacy

Purchasing tokens from these transfer stations compromises user data confidentiality. Every interaction passes through intermediary servers, posing a risk of misuse as companies may utilize this data for further training of their models.

Preventative Measures and Constant Adaptation

The Game of Cat and Mouse. Anthropic is aware of these ongoing issues and has implemented measures to combat them by blocking IP ranges linked to VPNs and data centers associated with these transfer stations. However, Chinese providers often employ elastic architectures that permit IP rotation, making the traffic appear legitimate.

In conclusion, while Claude’s unavailability in China creates barriers, local developers have ingeniously navigated these obstacles, leading to the emergence of an underground economy that poses both risks and unforeseen advantages. Developers must tread carefully in this marketplace, weighing the benefits against the potential downsides related to privacy and service quality.



General News – 2