Anthropic Exposes Chinese AI Labs Copying Claude at Scale 2026

Anthropic has exposed a large-scale operation in which Chinese AI labs copied Claude's capabilities through a technique called distillation. The company's latest threat intelligence report details how seven Chinese AI labs allegedly harvested data from Claude, with Alibaba (Qwen) alone conducting 151 million exchanges.

The report also highlights more unusual tactics: Moonshot (Kimi) and DeepSeek reportedly routed their own users' queries to Claude without disclosure. This means users interacting with those models may have unknowingly received answers generated by Anthropic's Claude.

How Chinese Labs Copied Claude Through Distillation

Distillation is a method where a smaller or less capable model is trained to mimic the outputs of a more advanced model. In this case, Chinese labs allegedly used Claude's responses to improve their own models. According to Anthropic, Alibaba's Qwen was the most active, with 151 million exchanges.

Moonshot and DeepSeek took a different approach. They reportedly used transfer stations—intermediary services outside China—to bypass Claude's geographic restrictions. When users asked Kimi certain questions, those queries were actually sent to Claude, and Claude's responses were returned as if generated by the Chinese model.

The scale is staggering: Moonshot allegedly performed this operation over 23 million times between May and July, while DeepSeek did so 12.1 million times in just two weeks. Users, of course, were unaware.

Why This Matters for AI Security and Trust

This incident raises serious concerns about AI security and transparency. When users interact with an AI model, they expect the responses to come from that model. If a model is secretly outsourcing queries to another AI, it undermines trust and could mislead users about the model's true capabilities.

For Anthropic, this is a direct violation of its terms of service. The company's threat report aims to shed light on such practices and protect its intellectual property. It also highlights the challenges of enforcing geographic restrictions in a globally connected world.

Moreover, distillation can be a double-edged sword. While it can help improve AI models, it can also be used to copy proprietary technology without permission. This case underscores the need for robust security measures and international cooperation to prevent misuse.

Key Takeaways from Anthropic's Threat Report

• Seven Chinese AI labs allegedly copied Claude's capabilities through distillation.
• Alibaba (Qwen) conducted 151 million exchanges with Claude.
• Moonshot (Kimi) used transfer stations to bypass restrictions, with over 23 million operations in three months.
• DeepSeek performed 12.1 million operations in just two weeks.
• Users were not informed that their queries were being routed to Claude.

Comparison of Chinese AI Labs' Activities

What to Watch Next in the AI Copying Controversy

Anthropic's report is likely to intensify the debate over AI ethics and intellectual property. It remains to be seen whether the company will take legal action or if regulators will step in. The incident also puts a spotlight on the practices of Chinese AI labs, which have been rapidly advancing in recent years. The background to this story is set out in China Denies Copying US AI Models, Warns of Retaliation. A related thread runs through AI Consciousness Debate: Why I Doubt Claude Is Truly Aware. Earlier coverage in OpenAI Ethically Hacked Using Anthropic's Claude Chatbot shows how the argument developed.

For users, the key lesson is to be aware of the models they interact with and demand transparency. As AI becomes more integrated into daily life, trust and accountability will be paramount. Anthropic's disclosure is a step toward greater transparency in the industry. A closer look at the same subject appears in Anthropic Bans Cruel Behavior Toward Claude AI Models. Anyone following the wider picture can start with Anthropic Faces Multibillion-Dollar Song Copyright Lawsuit. The same theme surfaces in How to Keep Your Claude Chats and Google Files Private, which takes a different angle.

We will continue to monitor developments in this story and provide updates as more information becomes available. Context on how this played out before sits in How to Keep Your Claude Chats and Google Files Private Online. A companion piece, Anthropic: Claude May Aid Biological Weapons Development, approaches the same subject from another angle.