Anthropic, the company behind the Claude chatbot, announced Monday that three Chinese artificial intelligence firms illegally extracted technology from their system to enhance their own AI models. This revelation comes just weeks after OpenAI made similar accusations against Chinese competitors.
According to Anthropic’s blog post, the Chinese companies DeepSeek, Moonshot, and MiniMax established approximately 24,000 fraudulent accounts to generate over 16 million interactions with Claude. This activity violated the company’s user agreements and geographic access limitations.
The firms employed a method known as “distillation,” where an established AI system evaluates responses from a developing model, essentially copying the knowledge and capabilities of the mature technology, Anthropic explained.
In a previous memo obtained by Reuters, OpenAI had alerted U.S. government officials that DeepSeek was specifically targeting ChatGPT and other leading American AI companies to duplicate their models for training purposes.
“These campaigns are growing in intensity and sophistication. The window to act is narrow, and the threat extends beyond any single company or region,” Anthropic stated in their disclosure.
Each Chinese company focused on different aspects of Claude’s capabilities. DeepSeek concentrated on reasoning abilities across various tasks and developing alternatives to policy-restricted queries. Moonshot pursued advanced reasoning, tool integration, coding, and data analysis features.
MiniMax targeted coding capabilities, tool usage, and system coordination. Anthropic discovered this particular operation while it was still ongoing, before MiniMax had launched the model they were developing.
“When we released a new model during MiniMax’s active campaign, they pivoted within 24 hours, redirecting nearly half their traffic to capture capabilities from our latest system,” the blog post said.
None of the three Chinese companies responded to requests for comment regarding these allegations.
