Anthropic alleged in a June 10 letter that operators affiliated with Alibaba and Alibaba Qwen used nearly 25,000 fraudulent accounts to generate more than 28.8 million Claude exchanges in an attempt to extract model capabilities. Reuters and UPI reported the allegations, while a June 25 U.S. House transcript shows l...
Alibaba-affiliated operators allegedly used fraudulent accounts to generate more than 28.8 million Claude exchanges, according to a June 10 letter from Anthropic.
Anthropic accused operators affiliated with Alibaba and Alibaba Qwen of illicitly accessing Claude in what Reuters reported the company described as its largest known distillation attack to date.
In the June 10 Anthropic letter, published as a PDF by Music Business Worldwide, the company said operators affiliated with Alibaba and Alibaba Qwen generated more than 28.8 million Claude exchanges through almost 25,000 fraudulent accounts between April 22 and June 5, 2026.
UPI also reported Anthropic’s allegation that Alibaba’s AI lab and affiliated entities used about 25,000 fake accounts to initiate almost 29 million Claude interactions over a roughly six-week period beginning April 22, 2026.
The allegation centers on “distillation,” a method in which outputs from one AI model are used to train or improve another model. Distillation can be used legitimately when model owners permit it, but Anthropic’s claim is that the activity involved deceptive access and unauthorized extraction of Claude capabilities.
Reuters reported on June 24 that Anthropic accused Alibaba of illicitly extracting Claude AI model capabilities, citing a letter seen by the news agency. Reuters said Anthropic characterized the incident as its largest known distillation attack to date.
The sources provided do not include a detailed public response from Alibaba to Anthropic’s specific allegations. As a result, the claims should be understood as Anthropic’s allegations as reported by Reuters and UPI and described in Anthropic’s June 10 letter.
The issue also surfaced in a June 25 U.S. House of Representatives transcript. In that hearing record, a member stated that news had broken that Alibaba targeted Anthropic with a distillation attack to steal Anthropic’s intellectual property. The transcript then records testimony discussing the allegation.
That congressional reference does not independently prove the allegations. It does show, however, that Anthropic’s claims quickly became part of U.S. policy discussion around AI security, intellectual property, and cross-border access to advanced models.
The case highlights a growing concern for frontier AI developers: access controls are becoming part of model security. If a rival can create large numbers of accounts and collect millions of model interactions, the provider may face risks that go beyond ordinary misuse. Those risks include unauthorized model imitation, leakage of proprietary behavior, and the creation of datasets designed to reproduce a model’s strengths.
Anthropic’s letter, as described in the cited sources, frames the incident as a deliberate campaign rather than routine user activity. The scale alleged — almost 25,000 accounts and more than 28.8 million exchanges — is central to the company’s claim that the activity was systematic.
For businesses using AI services, the dispute is also a reminder that model access, vendor policies, and compliance obligations are increasingly linked. Companies that build products on third-party AI systems may need to monitor not only what a model can do, but also how access is obtained, whether usage complies with provider terms, and whether outputs are being used to train competing systems.
The public record cited here establishes that Anthropic made the allegations and that Reuters, UPI, and a U.S. House transcript reported or discussed them. It does not establish, based on the provided sources alone, a final legal finding that Alibaba or any affiliated operator committed wrongdoing.
The next questions are factual and legal: how Anthropic attributed the accounts to Alibaba-affiliated operators, whether Alibaba disputes the attribution, and whether regulators or courts take up the matter. Until then, the case stands as a significant allegation in the widening debate over AI model access, data extraction, and intellectual property protection.
Alibaba affiliated operators allegedly used fraudulent accounts to generate more than 28.8 million Claude exchanges, according to a June 10 letter from Anthropic.
The allegation centers on “distillation,” a method in which outputs from one AI model are used to train or improve another model.
Distillation can be used legitimately when model owners permit it, but Anthropic’s claim is that the activity involved deceptive access and unauthorized extraction of Claude capabilities.
Continue reading
