The TL;DR
Anthropic accused Alibaba’s Qwen lab of using 25,000 fake accounts for nearly 29 million Claude transactions, the largest operation yet.
Anthropic accused Alibaba of carrying out a massive campaign to abort the US AI company, telling senators and White House officials that operators connected to Alibaba’s Qwen AI lab used 25,000 fraudulent accounts to de-authorize Claude between April and June. The book, a copy of which was seen by Bloomberg, described nearly 29 million exchanges with Claude that focused on software engineering and agent reasoning, the most commercially important modeling skills.
The lawsuit marks the first time Anthropic has named a major Chinese technology conference as the source of a distillation attack. Previous allegations in February were directed at Chinese AI startups, including DeepSeek, MiniMax, and Moonshot AI, which Anthropic said generated more than 16 million transactions using about 24,000 fake accounts. Alibaba’s campaign alone exceeded the combined volume of all three previous efforts.
Distillation is the practice of feeding carefully crafted questions to a frontier AI model, collecting its answers, and using those answers to train a cheaper competitor system that’s nearly as powerful as the original. The White House flagged the process as a national security concern in April, when OSTP Director Michael Kratsios published a memo obliging the government to share intelligence with US AI labs about foreign distillation campaigns. Anthropic said in its letter that the Alibaba campaign took place after the Kratsios memo, in defiance of the administration’s warnings.
💜 for EU tech
The latest talk from the EU tech scene, a story from our genius founder Boris, and some incredible AI art. It’s free, every week, in your inbox. Register now!
Alibaba did not comment on the allegations. An Anthropic spokesman declined to discuss specifics but stressed the importance of fighting distillation through joint action between government and industry.
Alibaba’s American depositary receipts fell more than 3 percent on the news, falling below $100 in afternoon trading Wednesday. The stock’s decline adds to the company’s difficult time in Washington, where it is facing pressure on multiple fronts.
The Pentagon added Alibaba to its blacklist of Chinese military companies on June 8, Anthropic said in its letter. Alibaba sued the Defense Department this week to win delisting, calling the label baseless and saying it had nothing to do with the military. The distillation case is now turning two, establishing Alibaba not just as a company with suspected military ties but as an active participant in what Anthropic calls the systematic theft of American AI skills.
In its letter, Anthropic warned that adversarial distillation allows Chinese labs to multiply the frontier of AI with minimal training costs, and that models built in this way often lack safety nets. The company urged the Trump administration to clarify antitrust guidelines so US labs can share more information about distillation efforts, reiterated its support for export controls on advanced AI chips, and called for fines against firms that use the method.
Lawmakers go hand in hand. Senators Bill Hagerty and Andy Kim plan to introduce an amendment to the defense law that must be passed that would blacklist or punish any Chinese company found to be improperly accessing US AI models. A bipartisan bill in the House, supported by representatives Bill Huizenga and Sydney Kamlager-Dove, is also being considered, although whether the proposal survives until the final version of the defense bill is uncertain.
Time is sensitive to Anthropic too. The company, now valued at $965bn after a Series H round of $65bn, filed privately for an IPO this month and is preparing for a listing that could come as soon as this fall. US officials have estimated that unlicensed distillation is costing Silicon Valley labs billions of dollars, and the threat of cheap imitation products from China poaching customers is the company’s biggest risk going public.
Anthropic’s requests for government support may not find a fully receptive audience, given that the company is facing a separate dispute with the Trump administration regarding export controls placed on its Fable 5 and Mythos 5 models less than two weeks ago. Commerce Secretary Howard Lutnick signed an order barring outsiders from accessing those models, citing safety concerns, and Anthropic failed to comply. Even after meetings between the company’s technical staff and White House officials, little progress has been made to restore services.
The result is a company caught between its two sides. Anthropic needs the government to strengthen the Chinese labs that export its technology, but at the same time it is fighting the same government decision to restrict its products. The letter to senators is an attempt to separate the two issues, arguing that protecting US models from distillation and allowing those models to be distributed commercially are complementary rather than conflicting goals.
Whether Washington agrees will change the regulatory landscape of US AI companies and the competitive landscape of the industry that follows. Anthropic has now named four Chinese labs as distillers of its technology, with Alibaba being accused of being too large. If the legislative proposals become stronger, the consequences could extend beyond the Anthropic models to the broader question of how the US applies the intellectual property boundary around AI systems that exist as software, not hardware, and can be copied over the Internet without anything more than well-designed information.
