Anthropic has accused Chinese tech giant Alibaba of orchestrating the largest known attack to steal capabilities from its Claude AI model. In a letter to U.S. senators, Anthropic presented evidence that between April 22 and June 5, 2026, operators linked to Alibaba and its AI lab Qwen used nearly 25,000 fraudulent accounts to generate over 28.8 million exchanges with Claude.
The campaign targeted Claude's most valuable features, including agentic reasoning, software engineering, and long-horizon tasks, violating terms of service and access restrictions. Anthropic is calling for punishment, arguing the effort was a coordinated attempt to replicate its leading model amid China's push to close the AI gap.
Comments