Despite strict US export controls and Anthropic’s rigorous geo-blocking measures, Chinese developers continue to access banned American AI models like Claude through a well-established proxy network known as the "transfer station" economy. This market allows broad participation beyond elite labs, highlighting shortcomings in current AI access restrictions and introducing new risks.
- Proxy-based transfer stations circumvent US AI export restrictions in China
- Access involves widespread users beyond elite AI research labs
- Proxy market challenges AI safety, traceability, and governance frameworks
What happened
In 2026, Chinese entities have developed an extensive grey market to access US AI frontier models like Anthropic’s Claude, which are officially banned under US export controls. This market, known as the "transfer station" economy, operates through multiple layers of API proxies that allow users across China—ranging from individual developers and students to tech workers—to use these advanced AI tools at prices as low as 10% of the official cost. Despite Anthropic’s rigorous geo-restriction techniques including geoblocking, mandatory overseas phone numbers and credit cards, and biometric KYC checks, the transfer stations effectively mask user identities and locations to bypass these controls.
This proxy-driven ecosystem is publicly visible on platforms such as GitHub, Taobao, Twitter, and Telegram, illustrating its broad reach and normalcy within the Chinese internet landscape. It supports tens of thousands of accounts acting as middlemen between US AI providers and local users, rendering official attempts to block or monitor access largely ineffective.
Why it matters
The transfer station economy demonstrates significant limitations in current AI governance methods that rely on blocking and monitoring user accounts to enforce export controls. The widespread use of proxies not only nullifies geographic restrictions but also enables anonymous access and usage data collection that can be bought and sold within China. This undermines the traceability intended by US AI companies to prevent misuse and enforce safety frameworks, exposing blind spots that extend beyond geopolitical considerations.
Moreover, the layered evasion infrastructure supporting these proxies introduces additional risks. With the rise of SMS farms and biometric data harvesting, the supply chain extends into illicit areas that may exploit vulnerable individuals. The transfer station economy thus poses challenges for US providers and regulators seeking to prevent harmful applications of AI while protecting individual privacy and maintaining control over their technologies.
What to watch next
The ongoing evolution of the transfer station ecosystem will be critical to monitor as Chinese users grow more sophisticated in circumventing AI export restrictions. Key indicators will include the scale and sophistication of proxy networks and the emergence of new evasion tactics such as enhanced biometric spoofing or fake identity verification. These developments could further complicate enforcement efforts by US AI firms and regulators.
At the same time, the transfer station model highlights the need for AI safety frameworks and export controls to adapt beyond simple access blocking. Future policies may require stronger cooperation between AI providers, governments, and international partners to address supply chain vulnerabilities, data commodity markets, and broader societal risks. How these dynamics play out could shape the future of AI governance and cross-border technology regulation.