Anthropic alleges Alibaba conducted largest Claude cloning attack on record

Anthropic has escalated accusations against Chinese tech giant Alibaba, claiming the company orchestrated the largest-ever attempt to reverse-engineer Claude through mass data extraction.

In a letter to Senators Tim Scott and Elizabeth Warren ahead of a June 11 Senate hearing on AI, Anthropic disclosed "new, confidential evidence" of what it characterized as an unprecedented distillation campaign. Between April 22 and June 5, operators affiliated with Alibaba and its AI division Qwen allegedly generated "more than 28.8 million exchanges with Claude through almost 25,000 fraudulent accounts," the company said.

The attack targeted Claude's most advanced features—agentic reasoning, software engineering, and long-horizon task planning—while exploiting obfuscation techniques and proxy networks to avoid detection. Anthropic warned that such circumvention tactics are spawning a "growing circumvention economy" that enables repeated extraction attempts.

Timing raises geopolitical stakes#

The timing adds political weight to Anthropic's claim. The alleged attack continued after the Trump administration took action in April to curb AI model theft, with the US accusing China of "industrial-scale" AI theft. That same month, Anthropic had publicly accused DeepSeek, Moonshot, and MiniMax of similar tactics—generating over 16 million queries through roughly 24,000 fake accounts.

Anthropic argues these distillation campaigns effectively "turn hundreds of billions of dollars in American investment and R&D into a massive subsidy for our geopolitical competitors." OpenAI and Google have reported finding similar attacks on their models, suggesting the practice is widespread across frontier AI labs.

The company has called for Alibaba to face unspecified punishment, framing the breach as a national security concern and a violation of Claude's terms of service.