Anthropic, one of America’s most valuable artificial intelligence firms, has accused the Chinese e-commerce and technology giant Alibaba of “brazenly” and “illicitly” extracting the capabilities of its Claude AI model, in what it has branded the largest campaign of its kind yet seen.
In a letter to senior members of the US Senate Banking Committee, the San Francisco-based developer said operators linked to Alibaba conducted almost 29 million exchanges with Claude using roughly 25,000 fraudulent accounts. The activity, it said, ran between 22 April and 5 June and amounted to “the largest campaign to illicitly extract Claude’s capabilities” recorded to date, according to the company’s account first reported by CNBC.
The letter, addressed to committee chairman Tim Scott and ranking member Elizabeth Warren, urged Congress to penalise the companies behind such attacks and to tighten the measures designed to stop American technology being siphoned off by overseas rivals.
According to Anthropic, the operation relied on what are known as “distillation attacks”, a technique in which answers are extracted from a stronger AI model to train a weaker one, sidestepping the export controls that govern the sale of model weights themselves.
The Alibaba-linked operators are said to have targeted Claude’s most commercially valuable functions, among them agentic reasoning, software engineering proficiency and the ability to see longer, more complex tasks through to completion. Attacks of this kind, Anthropic argued, are now being run on an “industrial scale” so that Chinese firms can harvest American AI capabilities and repackage them as their own.
Support authors and subscribe to content
This is premium stuff. Subscribe to read the entire article.
