The American company behind the Claude family of computer programmes has accused Chinese technology giant Alibaba of conducting the largest known operation of its kind against Anthropic, using tens of thousands of fraudulent accounts over a period of 45 days to systematically copy the capabilities of software that took years and billions of dollars to build.
Anthropic set out its accusations in a formal letter addressed to the United States Senate Committee on Banking, Housing, and Urban Affairs, as well as to officials at the White House. The letter, which Reuters saw, describes an operation it says was carried out by parties affiliated with Alibaba and with Alibaba’s research laboratory, known as Qwen. Alibaba had not responded to requests for comment at the time of publication.
The operation ran between the 22nd of April and the 5th of June this year. During that period, Anthropic says the parties involved created nearly 25,000 fraudulent accounts and used them to conduct more than 28.8 million interactions with Claude. The scale of what is described is considerable. To put the figure in context, earlier this year Anthropic disclosed that a similar operation attributed to another Chinese company, MiniMax, had involved 13 million interactions. The Alibaba operation is described as more than twice that size and is what Anthropic calls the largest incident of this type it has encountered.
What was actually happening during those 28.8 million exchanges requires some explanation. The method being described is sometimes called distillation, a term that in this context has nothing to do with chemistry. It refers to a process in which one piece of software is used as a teacher for another. If you ask a highly capable programme thousands of questions and record all of its answers, you can use those answers to train a weaker programme to behave in ways it could not manage before, picking up patterns and capabilities from the stronger one without having to do the underlying research and development work that made the stronger one capable in the first place.
Anthropic describes this method in its letter as a way for competitors to build systems that approach the capabilities of leading American programmes at a fraction of the cost. “These attacks are carried out illicitly, systematically, and at an industrial scale to harvest US capabilities across frontier labs and repackage them as their own without incurring the training and research and development costs required,” the company wrote. It also stated that the safety protections and ethical guardrails built into legitimate programmes are typically stripped away during this process, meaning the resulting systems carry the underlying capability without the protections designed to prevent misuse.
The letter specifically states that the Alibaba operation was designed to help accelerate China’s ability to reach the level of capability represented by Anthropic’s most advanced programme, known as Mythos Preview, which is not available to the general public and is currently accessible only to a small number of vetted organisations as part of a controlled programme.
This is not the first time Anthropic has made accusations of this kind against Chinese companies. In February this year, the company publicly disclosed that it had identified similar operations attributed to three other Chinese technology firms. DeepSeek, which attracted enormous international attention in early 2025 when it released a capable programme at unexpectedly low cost, was said to have been involved in an operation involving more than 150,000 interactions with Claude. Moonshot AI was linked to a campaign at a scale of more than 3.4 million interactions, and MiniMax to one exceeding 13 million. At the time, Anthropic warned that these operations were growing in intensity and sophistication and called for coordinated action involving both technology companies and governments.
The political context surrounding the Alibaba accusation is unusually charged. Earlier in June, the United States Department of Defense added Alibaba to its list of companies considered to have ties to the Chinese military, a designation the company is contesting in the courts. The Commerce Department, meanwhile, has not yet placed the Chinese company DeepSeek on a formal trade restriction list, despite an intergovernmental committee reportedly having assessed it as a national security concern.
Anthropic is using its communications with Congress and the White House to make a broader argument about policy, pressing for greater sharing of intelligence between the government and private technology companies about threats of this kind, and for faster coordinated responses when such operations are detected. The company argues that the current pace of regulatory action does not match the pace at which these operations are being carried out.
Alibaba and its Qwen laboratory have been active in the field of large language programmes, releasing a series of models that have attracted considerable interest for their performance relative to their cost. Whether the company responds publicly to Anthropic’s accusations, and what it says when it does, will be watched closely across both the technology industry and the governments currently debating how to manage the increasingly fraught competition between American and Chinese firms in this field.
About the Author
Marcel Chidozie is a tech analyst and writer covering foreign news, fintech, and emerging technologies at TechRegard. Based in Nigeria, He's passionate about translating complex tech developments into compelling, accessible stories for diverse audiences. His work focuses on how technology shapes innovation across Africa and globally.
