Anthropic has accused operators linked to Alibaba’s Qwen AI lab of conducting the largest known distillation attack on its Claude artificial intelligence model, according to a letter the company sent to US senators and White House officials on June 10. The campaign involved nearly 25,000 fraudulent accounts generating more than 28.8 million exchanges with Claude between April 22 and June 5, Reuters reported. Anthropic described the effort as brazen and illicit, targeting the model’s most prized capabilities including complex reasoning, software engineering and agentic decision-making.
The letter, seen by Reuters, Bloomberg and the BBC, framed distillation attacks as a process that uses outputs from a stronger model like Claude to train weaker ones at far lower cost. Bloomberg reported that the operation undermined Anthropic’s restrictions preventing its technology from reaching the Chinese market. Such campaigns occur on an industrial scale to harvest and repackage US AI capabilities as their own, according to the document.
Anthropic stated in the letter that “Distillation attacks turn hundreds of billions of dollars in American investment and research and development into a massive subsidy for our geopolitical competitors.” The company cited US Department of Defense assessments that link Alibaba, BYD and Baidu to the Chinese military, though those firms have denied the allegations, Reuters noted. Alibaba sued the US government this week seeking removal from a Pentagon blacklist, the BBC reported.
This accusation follows Anthropic’s February report that identified distillation activities by several Chinese AI laboratories, including efforts to improve their own models using Claude outputs, according to a company publication reviewed by Reuters. OpenAI has made similar claims against Chinese entities in prior instances, industry coverage shows. The interconnect.substack.com analysis of the February findings highlighted that labs tied to major tech firms such as Alibaba’s Qwen pursue these methods alongside independent developers.
The letter urged Congress to penalize companies behind such attacks and strengthen safeguards against the theft of US technology that could threaten military applications, Bloomberg reported. Anthropic has geared up for a potential initial public offering that could rank it among the world’s most valuable companies while some of its advanced models including Mythos have drawn cybersecurity concerns, the BBC noted. Nikkei Asia reported that leading Chinese firms are developing counterparts to Anthropic’s Mythos 5 model amid the ongoing AI competition.
US officials have expanded export controls on frontier AI systems to China over national security risks, a policy context that Anthropic’s letter referenced as justification for tighter measures. The scale of the alleged Alibaba-linked campaign exceeded previous known distillation incidents detailed in Anthropic’s earlier disclosures. No direct response from Alibaba to the latest specific claims appeared in reporting from Reuters, Bloomberg or the BBC.
