Anthropic says Chinese AI firms stole Claude data, sparking U.S.-China AI feud
Anthropic claims three Chinese AI firms—DeepSeek, Moonshot AI and MiniMax—used roughly 24,000 fraudulent accounts to generate over 16 million Claude interactions, extracting the model’s capabilities through “distillation.” The alleged breach of Anthropic’s service rules highlights a new front in U.S.–China AI competition and raises security alarms.