Anthropic accuses Chinese AI labs of mining Claude

Anthropic accuses three Chinese AI firms — DeepSeek, Moonshot AI and MiniMax — of creating more than 24,000 fake accounts to generate over 16 million exchanges with its Claude model. The company says the labs used a technique called “distillation” to extract Claude’s most differentiated capabilities: agentic reasoning, tool use and coding.

Anthropic traced different campaigns at each lab. DeepSeek logged more than 150,000 exchanges that seemed aimed at foundational logic and alignment, including censorship‑safe alternatives to policy‑sensitive queries. Moonshot generated over 3.4 million exchanges targeting agentic reasoning, tool use, coding, data analysis, computer‑use agent development and computer vision; it recently released an open source model, Kimi K2.5, and a coding agent.

MiniMax conducted roughly 13 million exchanges focused on agentic coding, tool use and orchestration, and Anthropic says MiniMax redirected nearly half its traffic to siphon capabilities from the latest Claude launch.

China

anthropic, claude, deepseek, moonshot ai, minimax, model distillation, model extraction, fake accounts, kimi k2.5, coding agent