Anthropic has accused Chinese firm Alibaba of orchestrating the largest known campaign to illicitly extract capabilities from its Claude AI model, sending a letter to U.S. senators on June 10 with new evidence.
The attacks, which occurred between April 22 and June 5, involved over 28.8 million exchanges made through nearly 25,000 fraudulent accounts operated by affiliates of Alibaba and its AI lab, Qwen. Anthropic stated the campaign targeted Claude's most valuable features, including agentic reasoning and software engineering, following the release of its Mythos model and subsequent restrictions on foreign access.
The company is calling for punishment against Alibaba for violating Claude's terms of service and U.S. market limitations.
Comments