Anthropic Claims Alibaba Ran 'Brazen' Campaign to Access Its Claude AI Model — Update

By Jason Chau and Tracy Qu

U.S. artificial-intelligence lab Anthropic has accused Chinese e-commerce giant Alibaba of "illicitly" accessing its frontier model Claude to harvest its capabilities.

In a letter obtained by The Wall Street Journal, Anthropic alleged that Alibaba and its AI unit conducted "the largest known distillation attack" on the company to date.

Distillation attacks, which involve the illicit training of a less capable model on the outputs of a stronger one, are turning "billions of dollars in American investment and R&D into a massive subsidy for our geopolitical competitors," it said.

The letter, dated June 10, was addressed to U.S. Senators Tim Scott and Elizabeth Warren, and claimed that Alibaba created almost 25,000 fake accounts to access Claude through nearly 29 million exchanges with the AI model, which is not available to entities in China.

According to the letter, "Alibaba's campaign targeted some of Claude's most valuable capabilities, such as agentic reasoning, software engineering, and long-horizon tasks."

An Anthropic spokesperson declined to comment on the contents of the letter but said: "We believe combating the threat of illicit distillation requires coordinated action between government and industry, and we will continue working with Congress and the Administration to maintain American AI leadership."

Alibaba didn't immediately respond to a request for comment.

It's not the first time the Anthropic has said Chinese AI labs are using its technology to train their own AI models.

In a blog post in February, Anthropic said it had identified industrial-scale campaigns by three AI laboratories to extract Claude's capabilities.

The latest accusations come after the Trump administration issued a memorandum in April saying that foreign entities, principally based in China, are engaged in industrial-scale campaigns to distill U.S. frontier AI systems.

Anthropic's letter said the company is supportive of Washington's efforts to combat such attacks, while also calling for more action to secure American AI, including tightening chip controls and legislation to penalize AI labs found to be engaging in distillation.

Shares of Alibaba dropped 4.4% in Hong Kong on Thursday, underperforming the Hang Seng Tech Index's 1.6% decline. But that came alongside a broader decline in other China tech majors trading in Hong Kong.

Investors don't appear concerned by Anthropic's accusation, Nomura analyst Jialong Shi said in an email.

"Given this is not the first distillation allegation targeting Chinese firms, I expect the reputational damage ... to be small," said Laila Khawaja, a research director at Gavekal Technologies.

What matters at this point is whether Anthropic will take further measures to crack down on distillation, and whether its repeated complaints to the U.S. government will lead to concrete export controls, she said.

However, there are substantial challenges to achieving both those outcomes, Khawaja added. For one, it is "extremely difficult" for AI labs to distinguish between real and fake requests, she said, and stepping up restrictions on Chinese AI companies would likely require more evidence and enforcement.

The claims against Alibaba come as Beijing and Washington step up their bids for AI dominance.

Anthropic's models still lead the global rankings for performance, though Chinese counterparts are gradually closing the gap. Beijing-based Zhipu AI's latest model is fourth in the global ranking compiled by the Artificial Analysis Intelligence Index, while Alibaba's Qwen 3.7 Max model placed eighth.

Write to Jason Chau at jason.chau@wsj.com and Tracy Qu tracy.qu@wsj.com

(END) Dow Jones Newswires

June 25, 2026 05:08 ET (09:08 GMT)