Anthropic has told US lawmakers that operators tied to Alibaba and its Qwen lab ran nearly 25,000 fraudulent accounts to pull capabilities out of Claude, generating 28.8 million exchanges between April 22 and June 5. The accusation came in a June 10 letter to the Senate Banking Committee, addressed to Chair Tim Scott and ranking member Elizabeth Warren, and was first reported by Bloomberg before CNBC confirmed a copy.
The number that matters
28.8 million. That is the figure Anthropic put on the campaign, and it is the one worth sitting with, because the company's own earlier disclosures give it context. Back in February, Anthropic said DeepSeek, Moonshot, and MiniMax had collectively run about 16 million exchanges. So one operation, if the attribution holds, beats all three of those combined by a wide margin.
How does Anthropic know it was Alibaba? The company says it ties accounts to specific actors through IP patterns, request metadata, and infrastructure fingerprints. That is the same playbook it described in its February distillation post, and it is worth remembering that this is a vendor's attribution of activity on its own platform. Alibaba has not commented. Take the precision of "28.8 million" as Anthropic's count, not an audited one.
What they were after
Software engineering and agentic reasoning. Anthropic frames these as Claude's most commercially valuable capabilities, which is true, and also exactly what you would expect a company to say about the thing it claims someone tried to steal. The technique is adversarial distillation: prompt a stronger model at scale, collect the outputs, train a cheaper rival on them. No source code required.
"These distillation attacks are carried out illicitly, systematically, and at industrial scale to harvest US AI capabilities across frontier labs and repackage them as their own," Anthropic wrote. Strong language for a letter that also happens to be a policy ask.
And it is a policy ask. Anthropic wants Congress to clarify antitrust rules so US labs can share threat intelligence on distillation more freely, and it wants penalties for firms doing it. The company, valued at $965 billion in private markets and prepping an IPO, has an obvious commercial stake in cheaper Chinese imitators not eating its lunch. That does not make the accusation wrong. It does mean the framing is doing double duty.
The awkward timing
Two days after Anthropic sent the letter, on June 12, the Commerce Department slapped export restrictions on its newest models, Fable 5 and Mythos 5, blocking access by any foreign national, including Anthropic's own non-citizen employees. So the same week Anthropic asked Washington to help it fight Chinese extraction, Washington restricted who Anthropic could serve. The company disabled access to those models while it works to comply.
Alibaba is fighting on a separate front too. The Pentagon added it to a blacklist of firms alleged to support China's military earlier this month, a designation the company is contesting in federal court. Anthropic's letter cited that listing directly.
A Senate amendment from Tennessee's Bill Hagerty and New Jersey's Andy Kim would blacklist or sanction Chinese firms found improperly accessing US model output. Whether it survives into the final defense bill is unsettled. That vote is the next thing to watch.
