Anthropic Says Chinese AI Firms Used 16 Million Claude Queries to Copy Model

Anthropic has uncovered large-scale illicit campaigns by AI companies DeepSeek, Moonshot AI, and MiniMax, aimed at extracting proprietary capabilities from its Claude large language model (LLM). These campaigns involved over 16 million interactions via approximately 24,000 fraudulent accounts, constituting a significant violation of terms and highlighting ongoing challenges in protecting AI models from unauthorized data extraction and model distillation attacks.