Anthropic told the Senate Banking Committee that operators affiliated with Alibaba’s Qwen AI lab ran roughly 25,000 fraudulent accounts through more than 28.8 million exchanges with Claude between April 22 and June 5, in what the company calls the largest known distillation attack on its models. The accusation, contained in a June 10 letter to Chair Tim Scott and Ranking Member Elizabeth Warren and made public June 24, frames the campaign as targeting Claude’s software engineering and agentic reasoning capabilities, the parts of the model Anthropic considers most commercially valuable.

The letter, addressed also to Senators Bill Hagerty and Andy Kim and Representatives Bill Huizenga and Sydney Kamlager-Dove, places the Qwen volume against a February incident in which Anthropic attributed roughly 16 million exchanges across about 24,000 accounts to three smaller Chinese labs: DeepSeek, MiniMax and Moonshot AI. The Alibaba campaign, on Anthropic’s numbers, eclipses that combined effort by nearly two-to-one in exchanges.

What gives the disclosure its political edge is timing. Bloomberg, citing the letter, reports the campaign ran “in defiance” of an April OSTP memo from Director Michael Kratsios committing the federal government to share foreign-distillation intelligence with U.S. labs. Two days after the letter arrived on the Hill, on June 12, the Commerce Department ordered Anthropic to suspend foreign-national access to its Fable 5 and Mythos 5 models. Alibaba’s U.S.-listed shares fell more than 3 percent, dropping below $100 in afternoon trading.

Distillation has always been the quiet backchannel of the frontier-model economy: capabilities flow from whoever pays the compute bill to whoever queries the API. Anthropic is now asking Washington to treat that backchannel as an export-control problem, and Commerce has already answered.

Sources