The US artificial intelligence firm Anthropic has accused Alibaba, a Chinese e-commerce and technology giant, of brazenly extracting the capabilities of its Claude AI model. The alleged operation, which took place between 22 April and 5 June, involved almost 29 million exchanges using roughly 25,000 fraudulent accounts.
Anthropic stated in a letter to senior members of the US Senate Banking Committee that the activity amounted to the largest campaign to illicitly extract Claude's capabilities recorded to date. The company urged Congress to penalize the companies behind such attacks and to tighten measures designed to stop American technology being siphoned off by overseas rivals.
The operation relied on 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. Alibaba-linked operators targeted Claude's most commercially valuable functions, including agentic reasoning, software engineering proficiency, and the ability to see longer, more complex tasks through to completion.
Anthropic argued that attacks of this kind are now being run on an industrial scale, allowing Chinese firms to harvest American AI capabilities and repackage them as their own. The financial stakes are considerable, with the company stating that distillation attacks turn hundreds of billions of dollars in American investment and research and development into a massive subsidy for geopolitical competitors.
This is not the first time Anthropic has raised the alarm about distillation attacks. In February, the firm said it had identified three separate industrial-scale distillation campaigns linked to Chinese labs. The Alibaba episode dwarfs all three, with its figures indicating a significant escalation of the problem.
The letter also pointed to alleged activity that Anthropic said could threaten the US military, citing the Department of Defense's assessment that Alibaba has ties to China's armed forces. Alibaba has rejected any such suggestion and filed a lawsuit against the US government seeking removal from the Pentagon's 1260H list, which designates firms judged to be Chinese military companies.
American developers have repeatedly accused Chinese competitors of using distillation to build rival systems at a fraction of the cost of training a frontier model from scratch. The accusations land at a delicate juncture for Anthropic, which is being tipped for a stock market debut that could rank it among the most valuable businesses in the world.
The company's frontier technology has also become a lightning rod for security concerns, with its most advanced models alarming governments over their capacity to find and exploit weaknesses in computer systems. The dispute is a reminder that the technology underpinning productivity gains for small and medium-sized businesses is now bound up in a high-stakes contest between the world's two largest economies, one in which the rules are still being written.
