Anthropic has told the Senate Banking Committee that operators affiliated with Alibaba’s Qwen A.I. lab conducted 28.8 million exchanges with Claude through thousands of fraudulent accounts, in what the company calls the “largest known” distillation attack against its models. The June 10 letter, addressed to Senators Tim Scott and Elizabeth Warren and first reported by Bloomberg on Wednesday, frames the episode as the most extensive attempt yet by a Chinese firm to piggyback on an American frontier lab.

The accusation lands on a market already primed to read it as confirmation. Alibaba’s Hong Kong-listed shares fell as much as 4.9 percent Thursday, touching a 16-month low and extending the year’s decline to 33 percent. Xiaomi and Baidu each dropped more than 3 percent. A representative for Alibaba didn’t immediately respond to Bloomberg’s requests for comment.

Distillation, as Bloomberg’s companion explainer lays out, is the technique by which a competitor uses the outputs of a frontier model to train a cheaper imitator. It’s the kind of harm conventional enforcement is structurally ill-equipped to address: there’s no stolen file to subpoena, only patterns of API traffic that look, account by account, like ordinary use. Anthropic says the targeted capabilities were software engineering and agentic reasoning, the exact frontier where Claude’s commercial moat sits.

The political timing is the story underneath the story. Anthropic disclosed similar activity by DeepSeek, Moonshot, and MiniMax roughly four months ago. Two months ago, the White House Office of Science and Technology Policy issued a memorandum pledging to help A.I. companies detect distillation campaigns. Senators Bill Hagerty and Andy Kim are now moving to attach an amendment to pending defense legislation, per CNBC. The letter routes through Banking, not Commerce or Intelligence, which tells you which committee the company thinks holds the live pen.

Anthropic’s letter also cites Alibaba’s presence on the Pentagon’s list of Chinese military companies, a designation Alibaba is currently suing the Defense Department to escape. According to Nikkei Asia, the Trump administration recently ordered Anthropic to restrict access to its Mythos 5 frontier model from foreign nationals, including its own employees, citing unspecified “national security authorities.”

“We believe combating the threat of illicit distillation requires coordinated action between government and industry, and we will continue working with Congress and the Administration to maintain American A.I. leadership,” an Anthropic spokesperson said in a statement provided to Bloomberg and CNBC.

The phrase reads like compliance boilerplate, but it’s the operative ask. A private company is requesting that the U.S. government treat model weights the way it treats export-controlled semiconductors. Whether or not the 28.8 million number proves to be the right one, that regulatory frame is now in committee.

Sources