Chainalysis: The Shadow Market Behind China's AI Model "Distillation Tide" Bypassing U.S. Export Controls

Author: Chainalysis

Compiled by: Shenchao TechFlow

Senchao Insight: In the context of tightening global regulations on computing power and algorithms surrounding cutting-edge artificial intelligence, an underground proxy industry has emerged that specializes in reselling overseas large model interfaces to restricted regions. Chainalysis recently conducted field research and on-chain tracking of 100 AI proxy service providers.

The report indicates that about 40% of the proxy providers directly support cryptocurrency payments primarily in stablecoins and isolate the flow of funds through third-party escrow payment channels. On-chain monitoring also shows that whenever a new large model is released by leading U.S. companies, the payment frequency from related Chinese wallets experiences a significant spike.

Key Summary

• The shadow market is helping Chinese users bypass U.S. AI export controls. Proxy service providers offer access channels to restricted cutting-edge large models, enabling related companies to "distill" and integrate the capabilities of advanced U.S. models into their systems at a cost far lower than training from scratch.

• Cryptocurrency is driving the operation of this market. Chainalysis's research analyzes the evolving supply chain and payment infrastructure of the proxy ecosystem, revealing that 40% of the surveyed proxy service providers accept payments in cryptocurrencies primarily in stablecoins.

• Activity in Chinese-related wallets significantly spikes after the release of U.S. cutting-edge AI models. Data monitoring shows that whenever a new model is launched in the U.S., the transaction activity of suspected Chinese-related wallets is significantly higher than the normal baseline, indicating concentrated demand for the new model's capabilities, although on-chain aggregated data alone cannot directly confirm specific invocation motives.

Formation of Underground Access Channels for Overseas Large Models

The research and development of cutting-edge AI large models require substantial investment and are highly related to national security strategies. Based on this consideration, the U.S. government has explicitly prohibited users and companies in mainland China from accessing leading AI tools like Claude and ChatGPT to limit major competitors from acquiring related technologies.

However, Chainalysis's global intelligence team found that restricted Chinese-related users can still frequently access these restricted models. They rely on an underground proxy economy specifically designed to circumvent export controls, with many channels even reselling access at discounts below official prices.

Once access channels are obtained, relevant developers can extract outputs from top U.S. models at a very low cost, reinforcing their own model systems and saving substantial underlying R&D expenses.

These so-called "Proxy Services" act as intermediaries that route requests to overseas large models while concealing the true source of the requests. In terms of payment methods, they support local fiat channels like WeChat Pay and Alipay, as well as large-scale access to on-chain assets.

Chainalysis sampled 100 such AI proxy platforms and found that 40% also accept cryptocurrency payments, with the vast majority of settlement media being stablecoins.

Figure: Recharge interface of a surveyed AI proxy platform captured in August 2026, supporting payments via Alipay, WeChat, and TRC20, BSC, and Arbitrum network USDT.

Using cryptocurrency payments is likely a deliberate business strategy of these proxy services, as it can obscure the flow of funds and assist users in bypassing foreign exchange capital controls, providing "off-net" access channels for privacy-sensitive clients.

Chinese Companies Use Proxy Interfaces to Advance Model Training, Undermining Regulatory Effects

By providing indirect access to U.S. cutting-edge models, these proxy intermediaries are continuously undermining the access blockade set up by U.S. laboratories concerning national security.

With the help of "knowledge distillation" technology, Chinese AI laboratories can directly utilize the output responses of advanced large models to train and enhance their competitive models, at a cost far lower than building a training system from scratch.

Recently, Anthropic publicly accused three Chinese AI laboratories—DeepSeek, Moonshot, and MiniMax—of operating 24,000 fake accounts through proxy services to systematically extract Claude's model capabilities.

As Anthropic warned in its report, derivative models obtained through illegal distillation often strip away the built-in safety barriers of the original version, making it easy to spread high-risk technological capabilities. Just last month, the U.S. National Security Agency (NSA) also issued a formal warning, stating that such behavior poses significant risks to critical infrastructure, military, and technological economic security.

Once scattered and hidden gray channels have now evolved into a clearly structured commercial economy with well-defined upstream and downstream divisions of labor, establishing dedicated computing power account recharge and financial infrastructure.

In this emerging commercial network, the penetration of cryptocurrency payments is rapidly increasing, providing ready alternatives for clients looking to avoid domestic digital payment real-name tracking.

Suspicious Chinese-Related Wallet On-Chain Payments Surge After U.S. Model Releases

To verify the funding characteristics of this chain, we tracked the on-chain funding activities of suspected Chinese-related wallets before and after each major U.S. AI model release. If developers indeed rely on proxy services to rush to invoke new models and pay with cryptocurrency, there should be a noticeable transfer pulse after the new model is launched.

The statistical results confirmed this hypothesis: within the first three days after the release of U.S. cutting-edge AI models, the activity of suspected Chinese-related wallets transferring to the aforementioned proxy payment channels significantly surged, with the increase far exceeding the average of other wallets across the network.

Specifically, within the 0 to 2 days after the release of cutting-edge models, the payment activity of Chinese-related wallets was 55% higher than the normal baseline; however, around the 13th day after the model release (usually corresponding to the time window for the release of Chinese open-source weight models), their activity dropped to 21% below the baseline.

Figure: Changes in payment activity of suspected Chinese-related wallets every three days after the release of U.S. cutting-edge AI models (Source: Chainalysis)

This timing pattern aligns closely with the industry's urgent pace to seize and invoke the capabilities of new U.S. models, also confirming that crypto assets are acting as a foundational payment channel assisting model distillation. However, Chainalysis also pointed out that on macro aggregated transaction data alone, it is still not possible to directly pinpoint the exact invocation behavior behind each payment.

On-Chain Payments Often Transferred Through Escrow Channels to Proxy Providers

Geographically, the cryptocurrency wallets making payments to AI proxy providers are highly concentrated in mainland China, while also widely distributed in Japan, Hong Kong, Singapore, and South Korea, which aligns perfectly with the conventional paths through which Chinese proxy operators access overseas computing power via surrounding Asian corridors.

Further tracking revealed that the vast majority of AI proxy providers do not directly receive cryptocurrency payments from their own addresses but generally embed third-party escrow cryptocurrency payment processors.

This layered architecture provides threefold protection for the proxy ecosystem: first, it adds a buffer between the end buyer and the actual proxy operator; second, it helps evade banking cross-border foreign exchange regulatory restrictions; and third, it continues to maintain offshore concealed access channels for highly sensitive users.

Considering the strict ban on cryptocurrency payment transactions in mainland China, entrusting third-party payment service providers to handle payments can also significantly reduce the legal and operational impacts that proxy platforms would face from establishing their own collection nodes.

However, this nested payment channel also creates hidden compliance exposure for regulated compliant cryptocurrency exchanges. Because the funds of proxies that have been laundered through multiple layers of transfer ultimately still flow to compliant trading platforms, and it is difficult to directly identify their direct transactions with underground AI proxy providers on the surface ledger.

How Blockchain Intelligence Can Help Penetrate and Govern the Proxy Shadow Market

For U.S. law enforcement and regulatory agencies attempting to block the outflow of top models, AI proxy services present a severe enforcement challenge. When the underground resale ecosystem has already industrialized, simply blocking access to official websites at the source is no longer effective.

The multi-tiered supply and demand structure of the proxy industry chain adds to the difficulty of enforcement: upstream suppliers rely on account farming and model channel exchanges to obtain access; downstream proxy providers are responsible for aggregating computing power pools, forwarding requests, and selling subscriptions by volume. Even if one link in the channel is shut down, other suppliers can quickly step in to fill the gap.

In this context, blockchain intelligence tools can assist investigators in penetrating multiple layers of fog along the on-chain funding trajectory, accurately restoring the relationship map between underground proxy operators, paying users, escrow payment providers, and the trading platforms that ultimately receive the funds.

By accurately mapping these financial connections hidden in the dark, regulatory authorities and AI research institutions can identify key choke points in the industry chain and apply precise pressure, thereby systematically dismantling the underground funding network that sustains the illegal invocation of U.S. cutting-edge AI capabilities.

Frequently Asked Questions

What are Chinese AI proxy services?

They refer to commercial platforms that specialize in reselling access interfaces for large models like Claude, ChatGPT, and Gemini to restricted regions. Most proxies adopt a pay-as-you-go subscription model, allowing users to register for free and purchase token credits to initiate inference requests directly.

Why are regulators and AI companies highly alert to such services?

Beyond the ordinary usage needs of individual users, the greater concern lies in the potential for Chinese AI R&D teams to use this to extract the capabilities of U.S. cutting-edge large models in bulk to fine-tune and distill their own systems, thereby bypassing equivalent foundational R&D investments, circumventing export controls, and potentially producing derivative models lacking safety constraints.

To what extent do AI proxy services use cryptocurrency?

Cryptocurrencies represented by stablecoins allow proxies to operate outside the traditional banking system for payments, reducing payment friction and minimizing customer identity verification processes. Chainalysis's survey of 100 proxy platforms shows that 40% have integrated cryptocurrency payment functionality.

How do proxy service providers typically handle on-chain fund settlements?

Most service providers do not directly receive payments from backend wallets but heavily rely on third-party escrow cryptocurrency payment processors, thereby adding a buffer between buyers and operating entities, evading local financial regulatory restrictions and risks of fund tracking.