Anthropic has accused Alibaba of running the largest known campaign to clone its AI model's capabilities, and is pushing US lawmakers to respond.
In a June 10 letter to Senators Tim Scott and Elizabeth Warren — sent one day before a Senate hearing on AI — Anthropic described what it called "new, confidential evidence" of a coordinated extraction effort. Between April 22 and June 5, operators linked to Alibaba and its AI lab Alibaba Qwen allegedly created nearly 25,000 fraudulent accounts and generated more than 28.8 million exchanges with Claude. The campaign, Anthropic said, targeted specific high-value capabilities including agentic reasoning, software engineering, and long-horizon tasks.
The timing matters. Anthropic framed the attacks as a direct response to restrictions limiting foreign access to its leading model, suggesting Chinese AI developers are turning to systematic scraping when legitimate access is cut off. That framing makes this a policy story as much as a security one — Anthropic is essentially asking Congress to treat capability extraction as something worth legislating against.
Alibaba has not publicly responded to the accusations, and Anthropic's evidence remains confidential. It would not be the first time a major AI lab has alleged model cloning — Meta has previously flagged similar concerns — but 28.8 million exchanges across 25,000 fake accounts, if accurate, is a scale that makes prior incidents look like reconnaissance.