Anthropic Accuses Alibaba of Attempting to Steal Claude AI Secrets Using 25,000 Fake Accounts

Anthropic has launched a serious accusation against Chinese e-commerce giant Alibaba, alleging the company orchestrated the largest known 'distillation' attack on its artificial intelligence systems. The San Francisco-based firm claims that operators connected to Alibaba's Qwen AI division used nearly 25,000 fraudulent accounts to bombard its Claude chatbot with over 28 million queries between late April and early June 2026.

This development marks a significant escalation in the struggle for AI supremacy between the United States and China. According to a letter sent by Anthropic to Senators Tim Scott and Elizabeth Warren on 10 June, the campaign targeted high-value capabilities, including software engineering, long-horizon tasks, and agentic reasoning.

Alleged Operation Raises Alarm

The accusations against Alibaba are referred to as 'distillation,' a technique that relies on feeding a weaker AI model the outputs of a more powerful one, according to Anthropic. Distillation was used to speed up China's efforts to match the capabilities of Anthropic's Mythos Preview system. The AI company alleges that the campaign was run by individuals connected to Alibaba's Qwen AI division.

'Alibaba's campaign targeted some of Claude's most valuable capabilities, such as agentic reasoning, software engineering, and long-horizon tasks,' the letter reviewed by Reuters further stated.

According to the letter, which was addressed to Senators Tim ​Scott and Elizabeth Warren on 10 June 2026 ahead of an appointed date for a hearing on AI, distillation attacks turn 'billions of dollars in American investment and R&D into a massive subsidy for our geopolitical competitors.'

The reported scale, which Anthropic claims involved almost 29 million interactions, has led the company to conclude that the campaign was a coordinated effort rather than an isolated misuse by individual actors.

A Growing Pattern In The AI Arms Race

The allegations against Alibaba did not emerge in isolation. In February of this year, Anthropic accused several other Chinese AI companies of running campaigns involving millions of interactions and tens of thousands of accounts.

According to Anthropic, over 16 million interactions with Claude were generated by the Chinese AI startups DeepSeek, MiniMax, and Moonshot using over 20,000 accounts. These incidents also allegedly involved the use of distillation techniques to train competing models.

'These campaigns are growing in intensity and sophistication. The window to act is narrow, and the threat extends beyond any single company or region,' Anthropic stated in February, warning the public of the security risks from replicated systems that were missing critical safeguards designed to prevent misuse.

The latest claims are likely to amplify concerns in Washington about AI competition between the United States and China. Anthropic has urged policymakers to take stronger action against large-scale capability-extraction efforts, arguing that such activities could undermine American technological advantages.

A spokesperson for Anthropic said the company maintains that preventing the unauthorised replication of advanced AI systems cannot be achieved by the private sector alone and requires collaboration with government bodies, adding that the company plans to remain actively involved in discussions aimed at preserving the United States technological edge.

'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,' the spokesperson told The Wall Street Journal.

On 12 June, the US Commerce Department moved to restrict Anthropic's latest Mythos and Fable AI models, citing concerns that they could fall into the hands of China's military intelligence operations. This prompted Anthropic to shut down access to the models worldwide.

Alibaba has not publicly responded to the specific allegations regarding the Qwen division's involvement.