Anthropic blocks major Chinese AI firms' attempts to steal its tech via illicit training attacks

Anthropic has uncovered and stopped large-scale unauthorized attempts by Chinese AI companies including Alibaba, Moonshot, and DeepSeek to use its advanced AI model Claude to train their own systems. These "illicit distillation" attacks involved rerouting millions of customer interactions through Claude without consent, sometimes exposing sensitive information. This highlights growing tensions as US AI firms face intellectual property theft risks amid fierce global AI competition. Anthropic continues to monitor and defend its technology against such breaches.