Key Highlights
- Anthropic claims Alibaba executed an unprecedented AI “distillation attack” targeting the Claude platform
- Approximately 28.8 million interactions were conducted via roughly 25,000 illegitimate accounts from April 22 through June 5, 2026
- The operation sought to replicate Claude’s sophisticated features for training Alibaba’s proprietary AI systems
- Anthropic formally notified US Senators Elizabeth Warren and Tim Scott, calling for stricter enforcement measures
- Following the disclosure, the US Commerce Department imposed worldwide restrictions on Anthropic’s Mythos and Fable AI platforms
The San Francisco-headquartered artificial intelligence firm Anthropic has leveled serious allegations against Chinese technology conglomerate Alibaba, claiming the company orchestrated an extensive operation to duplicate its Claude AI system’s functionalities. According to Anthropic, Alibaba employed a technique known as “distillation attacks” to harvest sophisticated AI responses and leverage them for developing their own inferior models.
The alleged operation unfolded over a 44-day period spanning from late April to early June 2026. Throughout this window, entities connected to Alibaba and its artificial intelligence division, Alibaba Qwen, allegedly produced over 28.8 million conversational exchanges with Claude utilizing close to 25,000 counterfeit user accounts.
Anthropics characterizes this incident as the most extensive known assault of this nature directed at the organization.
Understanding Distillation Attacks
Distillation attacks function by submitting queries to a superior AI system and capturing those responses to enhance a less sophisticated model. This technique allows adversaries to replicate cutting-edge AI capabilities without investing in the costly research and development that created them.
Anthropic maintains that the perpetrators specifically targeted Claude’s most prized functionalities. These encompass its capacity to manage intricate, extended tasks and its sophisticated reasoning mechanisms.
The organization asserts these operations occur at an “industrial scale,” effectively transforming hundreds of billions of dollars in American artificial intelligence research into accessible resources for geopolitical competitors.
Appeal to US Lawmakers for Intervention
On June 10, Anthropic dispatched a formal correspondence to US Senators Elizabeth Warren and Tim Scott. Both legislators serve on the Senate Banking Committee, which had scheduled hearings concerning artificial intelligence around that period.
Within the communication, Anthropic advocated for Congress to impose sanctions on entities responsible for such operations and to reinforce safeguards protecting American AI innovations.
This represents not the initial instance of Anthropic raising concerns. Earlier in February, the company disclosed it had detected comparable campaigns orchestrated by Chinese AI companies DeepSeek, MiniMax, and Moonshot AI. The DeepSeek campaign encompassed more than 150,000 interactions, while MiniMax’s operation exceeded 13 million exchanges.
Alibaba has not issued any statement in response to inquiries. The corporation was recently added to the Pentagon’s registry of Chinese entities with military connections this month and is currently contesting that classification through legal channels.
The White House declared in April that China was engaging in industrial-scale theft of American AI intellectual property. OpenAI has similarly accused Chinese organizations of employing identical distillation methodologies in previous incidents.
On June 12, merely two days following Anthropic’s letter transmission, the US Commerce Department implemented limitations on Anthropic’s Mythos and Fable AI platforms. Department officials expressed concerns that these systems could potentially be exploited by military intelligence operations in China and other nations. Consequently, Anthropic suspended worldwide access to both platforms.


