Days after OpenAI warning, Anthropic accuses three Chinese AI labs of extracting its data

American artificial intelligence (AI) company Anthropic said three Chinese AI labs, DeepSeek, Moonshot AI and MiniMax, have extracted the capabilities of its Claude model to improve their own systems.

Claude is the company's AI assistant, designed to be helpful with a wide range of tasks, including coding, math, reasoning, research, and general conversation.

According to the San Francisco-headquartered Anthropic, the three labs generated more than 16 million interactions with Claude using roughly 24,000 fraudulent accounts, violating the company’s terms of service and regional restrictions.


The revelations come days after Anthropic’s rival OpenAI stated that DeepSeek and other Chinese outfits were taking vast amounts of data from ChatGPT, its chatbot. It warned that the Chinese startups were relying on new methods of distillation to tap into the technologies developed by American companies.

How they did it

In a blog post on Monday, Anthropic said the Chinese labs used a technique called “distillation.” It is a method that involves “training a less capable model on the outputs of a stronger one”. This makes models cheaper and faster to run.

However, Anthropic alleged that the competitors used distillation to copy Claude’s most advanced capabilities, including reasoning, coding, and tool use, without authorisation.

The company said it linked the activity to specific labs through IP tracking, metadata analysis, infrastructure patterns, and coordination signals. In some cases, Anthropic added that industry partners observed similar behavior on their platforms.

Also Read: Explained: What is Anthropic's AI tool that's sparking job loss fears

What they learnt

DeepSeek allegedly ran more than 150,000 exchanges, focusing on reasoning tasks and extracting step-by-step internal explanations.

Moonshot AI conducted over 3.4 million exchanges, targeting agentic reasoning, coding, and computer-use systems.

MiniMax generated more than 13 million exchanges, concentrating on coding and orchestration tools. Anthropic said MiniMax adjusted its activity within 24 hours of a new Claude model release.

Distillation disadvantage

Anthropic warned that models built through illicit distillation may lack safety safeguards designed to prevent misuse, including protections against bioweapons research, cyberattacks, and mass surveillance.

The company also argued that such extraction campaigns could undermine US export controls by allowing foreign labs to replicate advanced US AI systems without independently developing them.

Anthropic’s next steps

Anthropic said it is strengthening its defences on multiple fronts.

The Dario Amodei-led company is improving detection systems that identify suspicious usage patterns and coordinated account activity. It is also tightening account verification processes, particularly for high-risk access pathways.

Also Read: 100% of Claude's code effectively AI-written: Anthropic CPO