Anthropic accuses Alibaba of largest-ever attempt to extract Claude AI capabilities

In a serious escalation of tensions, US-based AI firm Anthropic has accused Alibaba, a leading Chinese tech conglomerate, of orchestrating an unprecedented attempt to illicitly extract capabilities from its Claude AI models. This allegation raises significant concerns in Washington regarding the transfer of advanced American AI technologies to China. The purported operation took place between April 22 and June 5, 2026, resulting in over 28.8 million interactions with Claude, facilitated through nearly 25,000 fake accounts. Anthropic characterized this effort as a 'distillation' operation, a technique where outputs from a more advanced AI system are utilized to enhance a smaller or less capable model. Anthropic claims that the campaign was executed by individuals linked to Alibaba and its AI research division, Alibaba Qwen. The company expressed fears that this operation could expedite China's ambition to emulate the sophisticated features of its own advanced Mythos Preview model. In a letter dated June 10, addressed to US Senators Tim Scott and Elizabeth Warren, the chair and ranking member of the Senate Banking Committee, Anthropic raised these concerns ahead of a forthcoming hearing on artificial intelligence. This revelation comes at a time when there is increased scrutiny in Washington regarding China's access to cutting-edge US AI technologies. Just recently, the White House accused China of engaging in large-scale intellectual property theft from American AI labs. Anthropic has voiced its support for US governmental measures to combat such threats, advocating for enhanced threat intelligence collaboration between the government and private AI enterprises. If validated, the scale of Alibaba's alleged operations would far surpass previous distillation attempts uncovered by Anthropic. Earlier this year, Anthropic reported on similar attempts by Chinese AI startup DeepSeek and two other laboratories, with DeepSeek's activities involving over 150,000 exchanges, Moonshot AI generating more than 3.4 million exchanges, and MiniMax exceeding 13 million exchanges. The company cautioned that these operations are becoming increasingly sophisticated and that addressing this issue demands prompt and coordinated actions from industry leaders and policymakers. Adding to Alibaba's challenges, the company has recently faced increased regulatory scrutiny in the US. It was included on the Pentagon's list of Chinese military companies, a designation that it is currently contesting. Meanwhile, the US Commerce Department has refrained from placing DeepSeek on a trade blacklist, despite a government committee's assessment that it poses a national security threat, likely to prevent further escalation of tensions with China. On June 12, just two days after Anthropic's letter, the Commerce Department imposed restrictions on Anthropic's latest Mythos and Fable models, citing concerns over their potential use by military and intelligence bodies in China and other nations of concern. These restrictions compelled Anthropic to globally disable access to these models, underscoring the mounting friction between US efforts to uphold technological superiority and the commercial aspirations of its prominent AI firms.