Key Points
- Anthropic has leveled allegations against DeepSeek, Moonshot, and MiniMax for conducting “distillation attacks” targeting Claude AI.
- The companies allegedly deployed 24,000 fraudulent accounts producing more than 16 million Claude interactions.
- Commercial proxy services were reportedly employed to circumvent Anthropic’s geographic restrictions blocking Claude access in China.
- MiniMax generated the highest traffic volume, responsible for more than 13 million of the total 16 million interactions.
- Anthropic advocates for collective action involving the AI industry, cloud service providers, and regulatory authorities.
Three Chinese artificial intelligence companies face allegations from Anthropic of orchestrating systematic campaigns designed to extract proprietary knowledge from Claude AI through a technique known as “distillation.”
We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax.
These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
— Anthropic (@AnthropicAI) February 23, 2026
DeepSeek, Moonshot AI, and MiniMax stand accused in the allegations. According to Anthropic, these operations spanned approximately 24,000 deceptive accounts and generated more than 16 million interactions with Claude.
Distillation represents a training methodology where a less capable AI model gains knowledge by analyzing outputs from a more advanced system. According to Anthropic, this approach serves legitimate purposes within organizations but crosses ethical boundaries when competitors deploy it to appropriate capabilities.
“Distillation can also be used for illicit purposes: competitors can use it to acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost,” Anthropic wrote in its blog post on Sunday.
Anthropic’s service terms prohibit commercial Claude access within China. According to the allegations, the three companies circumvented these restrictions by leveraging commercial proxy services, enabling simultaneous operation of tens of thousands of Claude accounts across various networks.
Following access, the firms reportedly transmitted substantial volumes of strategically designed prompts aimed at extracting specific capabilities from Claude. These responses subsequently served as training data for proprietary models and reinforcement learning applications.
The alleged attacks concentrated on Claude’s premier functionalities, encompassing agentic reasoning, tool utilization, programming, data analysis, and computer vision capabilities.
MiniMax Generated Highest Volume
Among the accused organizations, MiniMax generated the largest traffic volume. Anthropic reports that MiniMax independently produced more than 13 million of the cumulative 16 million interactions.
All three companies—DeepSeek, Moonshot AI, and MiniMax—maintain headquarters in China and hold multi-billion dollar valuations. The firms have remained silent on comment requests.
Anthropic identified the companies through multiple verification methods including IP address analysis, request metadata examination, infrastructure pattern recognition, and intelligence from industry collaborators who observed similar actors across different platforms.
Industry-Wide Concerns Emerge
Anthropic joins other US artificial intelligence companies addressing this concern. OpenAI submitted correspondence to US lawmakers recently, asserting observations of activity “indicative of ongoing attempts by DeepSeek to distill frontier models.”
OpenAI initially raised distillation concerns during early 2024, following the debut of DeepSeek’s inaugural model, which users observed bore striking similarities to ChatGPT.
Anthropic plans to strengthen detection capabilities, enhance access restrictions, and establish threat intelligence sharing partnerships with industry peers.
The company is also calling for a coordinated response from the wider AI industry, cloud providers, and policymakers. “No company can solve this alone,” Anthropic wrote.
Industry analysts consulted by CNBC observed that distinguishing between legitimate and improper distillation practices presents challenges, emphasizing the importance of careful evaluation when assessing such allegations.
Anthropic’s published statement confirmed that MiniMax generated over 13 million interactions, representing the largest volume attributed to any individual company in the alleged scheme.
