Crypto Payments Power a Shadow Market Giving Chinese Users Access to Restricted US AI Models
The US government bars users and companies in China from powerful American AI tools like Claude and ChatGPT because developing frontier AI models has national security implications. But new research from blockchain analysis firm Chainalysis shows a shadow market is helping suspected Chinese users get around those rules. The firm also found that crypto, mostly stablecoins — digital tokens designed to hold a steady value like the US dollar — plays a growing role in paying for this access.
Chainalysis's Global Intelligence Team surveyed 100 "proxy services." These are businesses that resell access to restricted AI models such as Claude, ChatGPT, and Gemini. They route queries into the models while hiding where the requests come from, often at a discount. Once inside, users can cheaply copy the advanced models' abilities to improve their own competing systems — a process called "distillation."
Key numbers and findings
- 40% of the 100 proxy services surveyed accepted cryptocurrency as payment, predominantly stablecoins.
- Payments from suspected China-linked wallets rose above normal levels in the days after new US AI model launches, suggesting efforts to quickly access new capabilities — though the aggregate data cannot confirm the cause.
- Most proxy services use custodial crypto payment processors — third parties that hold and forward funds — rather than collecting payments directly.
- Wallets paying for proxy services were heavily concentrated in China, as well as Japan, Hong Kong, Singapore, and South Korea.
What the Chainalysis research says
Proxy services take payment in regular money, including through China-based processors like WeChat and AliPay. Chainalysis says crypto payments are likely part of a deliberate strategy to obscure financial infrastructure, get around Chinese state capital controls, and keep off-grid access for sensitive users.
The researchers found that suspected China-linked wallets increased their activity in the three-day periods following frontier model releases, more than wallets overall using these services. Chainalysis notes this pattern is consistent with a push to gain fast access to new model capabilities, but the data cannot confirm what is driving it.
The firm says the market has matured from an informal trade into a structured economy with a segmented supply chain and dedicated financial infrastructure. Providers sell access through credits and account top-ups. Upstream vendors obtain model access through tactics like account farming and model swapping, while proxy services pool that access, route queries, and sell subscriptions. When one vendor is disrupted, others can step in.
Claims from Anthropic and the NSA
Anthropic recently accused three China-based AI labs — DeepSeek, Moonshot, and MiniMax — of extracting capabilities from Claude using 24,000 fraudulent accounts run through proxy services. These are the company's claims, not independently confirmed findings in the Chainalysis research.
Anthropic also says illicitly distilled models can spread dangerous capabilities because they often lack essential safeguards. Last month, the National Security Agency warned that China-based AI companies are distilling US frontier AI models and that such models can pose significant risks to critical infrastructure, military, economic, and technology domains.
What is confirmed and what is not
Confirmed by Chainalysis's own research: the survey of 100 proxy services and the 40% crypto acceptance rate, the use of custodial payment processors, the regional concentration of paying wallets, and the observed post-release activity spike among suspected China-linked wallets.
Reported but not confirmed by the data: that Chinese developers are specifically using crypto to pay for distillation after model launches. Chainalysis itself says the aggregate data cannot confirm the cause of the observed activity. The Anthropic accusations against the three Chinese labs are also claims by that company.
Why this matters for crypto companies and regulators
The layered payment structure creates exposure for regulated crypto exchanges and other crypto businesses. Chainalysis says funds linked to proxy services can reach regulated platforms with no clear direct ties back to the proxy services themselves.
China's broad restrictions on crypto payment activity may also push proxy services toward third-party processors, since that reduces their exposure to regulatory and operational risks of running their own payment infrastructure.
Chainalysis argues that blocking direct access to frontier models is no longer enough given the mature resale market, and that blockchain intelligence can help investigators trace financial flows, map connections between operators and the platforms where funds land, and identify pressure points to disrupt the networks.