Anthropic Accuses Alibaba of Largest Claude Distillation Attack

The part that makes Anthropic's accusation against Alibaba unusual isn't the allegation of copying; it's the alleged method and the scale at which it was apparently carried out. According to BBC News and multiple corroborating reports, Anthropic sent a letter to US Senators accusing operators linked to Alibaba's Qwen AI lab of using roughly 25,000 fraudulent accounts to run more than 28.8 million exchanges with Claude between April 22 and June 5, 2026. Anthropic described it as "the largest known distillation attack on Anthropic to date."

Model distillation, in this context, doesn't mean stealing source code or model weights. It means querying a frontier model at industrial scale, collecting its outputs, and using those outputs to train a competing system. The campaign reportedly targeted Claude's capabilities in software engineering and agentic reasoning, with the alleged goal of helping Alibaba's Qwen model approach the performance of Anthropic's Mythos Preview model, a system focused on coding, multi-step reasoning, and cybersecurity tasks.

This is not the first time Anthropic has named a Chinese AI lab in this kind of accusation. The company previously identified three earlier distillation campaigns attributed to DeepSeek, Moonshot AI, and MiniMax, which together reportedly generated over 16 million exchanges through about 24,000 fake accounts. The Alibaba campaign, by Anthropic's accounting, exceeded that combined volume on its own. Anthropic also alleged that the Chinese government was complicit in the effort, a claim with significant political weight but no independent public verification.

The honest caveat is that these are Anthropic's allegations and Alibaba has not publicly responded. What the reporting does not give you is independent technical confirmation that the fake accounts were actually linked to Alibaba or Qwen, or any evidence of how much the effort improved Qwen's performance. The US Commerce Department has since imposed export controls on Anthropic's Mythos and Fable models, and lawmakers are reportedly considering legislation targeting unauthorized AI model access, both of which suggest the government is treating the framing seriously, even if the specific evidentiary links remain one-sided for now.

For AI practitioners building on proprietary APIs, this case suggests that usage monitoring has become a competitive and national security concern, not just a billing problem.