Anthropic's letter reveals alleged Alibaba effort to distill Claude AI as US restricts AI access
On June 12, the US government asked Anthropic to suspend access to its advanced AI models, Mythos and Fable 5. While the move raised questions
On June 12, the US government asked Anthropic to suspend access to its advanced AI models, Mythos and Fable 5. While the move raised questions at the time, a newly revealed letter from Anthropic offers a glimpse into the concerns driving such decisions. According to the company, one of China's biggest technology firms may have been trying to gain access to its AI technology on a massive scale. Read Full Story Ahead of a scheduled hearing on artificial intelligence, Anthropic sent a letter, dated June 10, to Senators Tim Scott and Elizabeth Warren, the chair and ranking member of the US Senate Banking Committee. In the letter, Anthropic accused Alibaba Group of conducting a large-scale effort to "illicitly" access its Claude AI model by using thousands of fraudulent accounts. According to Anthropic, the campaign took place between April 22 and June 5, 2026, and involved nearly 25,000 fraudulent accounts that generated more than 28.8 million exchanges with Claude. The company alleged that the activity was carried out by operators affiliated with Alibaba and Alibaba Qwen, the company's AI research division.
Alibaba had not responded to requests for comment at the time the allegations became public. What is AI distillation? At the heart of the dispute is a practice known as "distillation." In simple terms, distillation involves training a less capable AI model using the responses generated by a more advanced one. Instead of building everything from scratch, developers can learn from the outputs of a stronger model and use that information to improve their own systems. Anthropic claims the alleged campaign was aimed at accelerating China's ability to reach capabilities similar to those of its advanced Mythos Preview model. The company views such efforts as a serious threat because they could help competitors replicate advanced AI capabilities without investing the same level of time and resources into research and development. Not the first accusation involving Chinese AI firms This is not the first time Anthropic has raised concerns about distillation. In February, the company said it had identified another campaign involving Chinese AI startup DeepSeek, along with two other Chinese AI labs.
According to Anthropic, DeepSeek's operation involved more than 150,000 exchanges with Claude. The company claimed Moonshot AI accounted for more than 3.4 million exchanges, while MiniMax generated over 13 million. At the time, Anthropic warned that such campaigns were increasing in both "intensity and sophistication" and argued that addressing the threat would require rapid and coordinated action across the AI industry and governments. Why the US is paying close attention Anthropic's latest allegations arrive at a time when Washington is becoming increasingly concerned about the flow of advanced AI technology. In its letter, the company expressed support for US government efforts to combat such activities through threat-intelligence sharing and cooperation with private AI companies. Alibaba was added to the Pentagon's list of Chinese military companies this month, a designation the company is challenging. At the same time, US officials have taken a more cautious approach toward other Chinese AI firms. DeepSeek, for example, has not been placed on a US trade blacklist despite reportedly being viewed as a national security risk by an interagency government committee.
The BriefWire