Anthropic accuses Chinese AI labs of mining Claude as US debates AI chip exports

Anthropic has publicly accused three leading Chinese AI companies—DeepSeek, Moonshot AI, and MiniMax—of orchestrating a coordinated 'distillation' campaign against its Claude AI model. The labs allegedly created over 24,000 fake accounts to generate more than 16 million exchanges with Claude, specifically targeting its most advanced capabilities: agentic reasoning, tool use, and coding. Distillation is a training technique where one model learns from another's outputs, but here it's being used competitively to extract proprietary capabilities.

The scale was massive: MiniMax alone generated approximately 13 million exchanges focusing on agentic coding and tool orchestration, while Moonshot AI produced 3.4 million exchanges targeting reasoning and computer vision. DeepSeek's 150,000 exchanges focused on foundational logic and alignment, particularly around censorship-safe alternatives. Anthropic observed MiniMax redirecting nearly half its traffic to target the latest Claude model upon launch. This follows OpenAI's recent memo to House lawmakers accusing DeepSeek of similar practices.

The timing is critical as the US debates AI chip export controls to China. Anthropic argues these distillation attacks 'require access to advanced chips' and reinforce the need for export restrictions. The Trump administration recently allowed exports of advanced chips like Nvidia's H200 to China, which critics say boosts China's AI capacity during a global race for dominance. Dmitri Alperovitch of Silverado Policy Accelerator stated this confirms suspicions about Chinese AI progress being fueled by 'theft via distillation of U.S. frontier models.'

Anthropic is calling for a coordinated industry, cloud provider, and policy response while investing in better defenses. The company warns that beyond commercial threat, distillation creates national security risks by potentially transferring restricted capabilities. This incident highlights the emerging front in AI competition where model outputs become intellectual property battlegrounds, with significant implications for both commercial strategy and geopolitical policy.