Anthropic has escalated tensions in the global AI race by accusing three Chinese artificial intelligence labs of systematically extracting capabilities from its flagship model, Claude. The Anthropic Claude distillation allegations claim that more than 24,000 fake accounts were used to generate over 16 million interactions in an effort to improve competing AI systems.

The accusations arrive as U.S. policymakers debate whether to tighten or loosen export controls on advanced AI chips to China.

Anthropic Claude Distillation Allegations Target Three AI Labs

According to Anthropic, the companies involved are DeepSeek, Moonshot AI, and MiniMax. The firm alleges that these labs used a technique known as “distillation” to mine Claude’s outputs and enhance their own models.

Distillation is a common AI training method used to create smaller or more efficient models from larger ones. However, when applied across competitors, it can effectively replicate a rival’s research and development.

Anthropic said the targeted capabilities included agentic reasoning, tool use, and coding — features that differentiate Claude from other models.

The scale of the activity varied by company. Anthropic reported more than 150,000 exchanges linked to DeepSeek that focused on foundational logic and alignment. Moonshot AI allegedly conducted over 3.4 million exchanges centered on agentic reasoning, coding, computer-use agents, and data analysis. MiniMax accounted for approximately 13 million exchanges targeting agentic coding and orchestration.

Anthropic stated that it observed MiniMax redirecting nearly half its traffic to siphon capabilities from the newest Claude model upon release.

AI Chip Export Controls in Focus

The Anthropic Claude distillation allegations come amid renewed debate over U.S. export controls on advanced AI chips. The policy is intended to slow China’s progress in AI development by restricting access to high-performance hardware.

Last month, the Trump administration formally allowed U.S. firms such as Nvidia to export advanced AI chips, including the H200 processor, to China. Critics argue that loosening these restrictions could accelerate China’s AI capabilities at a critical moment in the global competition for technological leadership.

Anthropic claims the scale of the alleged extraction activities would require access to advanced computing hardware, reinforcing its argument for maintaining export controls.

In a blog post, the company stated that restricting chip access limits both direct model training and the scale of illicit distillation.

Experts Weigh In on AI Model Extraction

Dmitri Alperovitch, chairman of the Silverado Policy Accelerator think tank and co-founder of CrowdStrike, told TechCrunch that such tactics have likely contributed to rapid advances in Chinese AI models.

The Silverado Policy Accelerator is a U.S.-based policy research organization focused on national security and economic resilience.

Alperovitch said the allegations strengthen the case for stricter AI chip export controls, arguing that continued hardware access would only enhance competitors’ capabilities.

Earlier this month, OpenAI also reportedly sent a memo to U.S. lawmakers accusing DeepSeek of using distillation techniques to mimic its models, further intensifying scrutiny of cross-border AI competition.

National Security and Safeguard Concerns

Beyond intellectual property concerns, Anthropic warned that distillation could undermine AI safety measures. The company stated that its models include safeguards designed to prevent misuse in areas such as biological research and malicious cyber activities.

According to Anthropic, models trained through illicit distillation may not retain these protections, increasing the risk of misuse by state or non-state actors.

The company also highlighted concerns that authoritarian governments could deploy advanced AI systems for offensive cyber operations, disinformation campaigns, and surveillance — risks that may multiply if such systems are open-sourced.

Anthropic said it will continue investing in defensive measures to detect and prevent distillation attacks while calling for coordinated action among AI developers, cloud providers, and policymakers.

The Anthropic Claude distillation allegations underscore how AI competition is increasingly intertwined with geopolitics, hardware supply chains, and national security strategy.