Anthropic: Alibaba Used 25,000 Fake Accounts to Copy AI Model

Anthropic accuses Alibaba's Qwen lab of orchestrating the largest known AI model copying operation. The AI firm told US senators that Alibaba used 25,000 fake accounts to generate over 28.8 million exchanges with its Claude system in just six weeks. This process, called distillation, involves a smaller model learning from a more powerful one. Anthropic claims the operation targeted software engineering and agentic reasoning functionalities of its Mythos Preview model. Alibaba has denied these accusations, calling them baseless. This isn't the first time Anthropic has raised concerns about Chinese AI labs copying its work. Previous campaigns were disclosed in February 2026, reportedly involving DeepSeek and Moonshot. Chinese models have accounted for a significant portion of token usage on platforms like OpenRouter. This issue has potential national security and intellectual property implications, and stricter export controls on AI model access are being discussed. For investors, if Chinese models capture a large share of developer consumption, it could significantly impact revenue for US AI companies.