Key Points
- Anthropic has publicly accused three Chinese AI companies — DeepSeek, Moonshot, and MiniMax — of conducting “distillation attacks” to replicate Claude AI’s capabilities.
- The alleged operation involved approximately 24,000 fraudulent accounts that facilitated more than 16 million interactions with Claude.
- Commercial proxy infrastructure was reportedly deployed to circumvent Anthropic’s geographical access limitations in China.
- MiniMax represented the largest volume contributor, responsible for more than 13 million of the total exchanges documented.
- Anthropic is urging industry-wide collaboration among AI developers, cloud infrastructure providers, and regulatory authorities to address this threat.
Anthropic has publicly named three Chinese artificial intelligence developers in what it describes as systematic efforts to replicate its Claude AI model’s capabilities through distillation techniques.
The companies identified in the allegations are DeepSeek, Moonshot AI, and MiniMax. According to Anthropic, the coordinated activity spanned approximately 24,000 fake user accounts and resulted in over 16 million Claude interactions.
Distillation refers to a methodology where less sophisticated AI systems are trained by analyzing the outputs generated by more advanced models. While Anthropic acknowledges this technique has legitimate internal applications, the company contends it becomes problematic when used by rivals to replicate proprietary technology.
“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 stated in a Sunday blog post.
Anthropic’s user agreement explicitly prohibits commercial Claude access from China. To circumvent these restrictions, the three companies allegedly leveraged commercial proxy networks to simultaneously operate thousands of Claude accounts.
Once access was established, the firms reportedly submitted high volumes of strategically designed prompts intended to extract specific functionalities from Claude. The resulting outputs were subsequently utilized for training proprietary models or as datasets for reinforcement learning processes.
The alleged extraction efforts targeted Claude’s most sophisticated capabilities, including autonomous reasoning, tool integration, software development, analytical processing, and visual understanding.
MiniMax Dominated Activity Volume
Among the three entities, MiniMax generated the highest interaction volume by a significant margin. Anthropic reports that MiniMax was responsible for over 13 million of the 16 million total exchanges identified.
DeepSeek, Moonshot AI, and MiniMax are all Chinese-based companies with valuations reaching into the billions of dollars. None of the three organizations provided responses to media inquiries.
Anthropic indicated it identified the companies through multiple verification methods including IP address analysis, request metadata examination, infrastructure fingerprinting, and intelligence shared by industry collaborators who observed similar patterns on their platforms.
A Broader Industry Concern
Anthropic is not the only American AI developer raising concerns about this practice. OpenAI submitted formal correspondence to United States legislators earlier this month, asserting it had detected activity “indicative of ongoing attempts by DeepSeek to distill frontier models.”
OpenAI initially highlighted distillation concerns in early 2024, following the release of DeepSeek’s inaugural model, which users observed bore striking similarities to ChatGPT.
Anthropic states it will strengthen its response through enhanced detection mechanisms, stricter access protocols, and expanded threat intelligence sharing with industry partners.
The company is also advocating for unified action across the AI sector, cloud service providers, and government entities. “No company can solve this alone,” Anthropic emphasized.
Industry analysts speaking with CNBC observed that distinguishing between acceptable and improper distillation practices can be complex, suggesting careful evaluation is necessary when assessing such allegations.
Anthropic’s official statement confirmed that MiniMax generated over 13 million interactions, representing the largest individual contribution among the accused firms.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
