February 24, 2026
Anthropic Accuses DeepSeek, Other China-Based AI Firms of Free-Riding
The company also expressed concerns about the recent U.S. approval of H200 chip exports to China—Nvidia's second-most advanced chip.
TL;DR
- Anthropic alleges that DeepSeek, Moonshot AI, and MiniMax created over 24,000 fraudulent accounts.
- These accounts reportedly sent over 16 million prompts to Anthropic's Claude chatbot.
- The purpose was to extract Claude's output for training the Chinese companies' own AI models via distillation.
- Anthropic described the tactic as 'distillation,' where a smaller model learns from a larger one.
- The company stated these deceptive campaigns are escalating in intensity and sophistication.
Continue reading the original article