Anthropic Accuses Alibaba of Creating 25000 Fake Accounts to Steal Claude AI Capabilities

Anthropic accuses Alibaba of orchestrating a massive AI distillation attack using 25,000 fake accounts to steal Claude's capabilities.

Jun 25, 2026
5 min read
Technobezz
Anthropic Accuses Alibaba of Creating 25000 Fake Accounts to Steal Claude AI Capabilities

Alibaba created nearly 25,000 fake accounts to pump Claude for its most valuable AI capabilities through almost 29 million exchanges, Anthropic alleges in a letter sent to U.S. senators and White House officials. The San Francisco AI lab called it the largest known distillation attack on the company.

Operators linked to Alibaba's Qwen AI lab targeted Claude's agentic reasoning, software engineering, and long-horizon task capabilities between April 22 and June 5, according to the letter sent to Senators Tim Scott and Elizabeth Warren. Bloomberg first reported the letter.

Distillation attacks involve training a weaker model on outputs from a stronger one, essentially stealing capabilities without permission. Anthropic said the technique is turning "billions of dollars in American investment and R&D into a massive subsidy for our geopolitical competitors."

Claude is not available to entities in China, and Anthropic said it had previously identified industrial-scale extraction campaigns by three AI laboratories in a blog post in February. The company urged Washington to tighten chip controls and pass legislation penalizing AI labs that engage in distillation. An Anthropic spokesperson declined to comment on the letter's specific contents but said the company is "working with Congress and the Administration to maintain American AI leadership."

Shares of Alibaba dropped 4.4% in Hong Kong trading Thursday, underperforming the Hang Seng Tech Index's 1.6% decline. Analysts said the market reaction was muted.

"Given this is not the first distillation allegation targeting Chinese firms, I expect the reputational damage to be small," Gavekal Technologies research director Laila Khawaja said. The accusation lands as Beijing and Washington escalate their competition for AI dominance. Anthropic's models still lead global performance rankings, but Chinese rivals are closing the gap.

Beijing-based Zhipu AI's latest model ranks fourth globally, and Alibaba's Qwen 3.7 Max sits at eighth, according to the Artificial Analysis Intelligence Index.

Share