Breaking

Anthropic Tells US Senators Alibaba Ran 28.8M-Exchange Claude Distillation Effort

June 27, 2026 at 00:01 EDT

Policy & Regulation
Security
Foundation Models

In a letter dated June 10, 2026, Anthropic told US senators that operators tied to China's Alibaba (and its Qwen AI lab) generated more than 28.8 million exchanges with Claude through roughly 25,000 fraudulent accounts, carrying out what it called the largest distillation effort yet to illicitly extract its model's capabilities. The company is asking Washington for tougher action against Chinese AI firms.

June 10, 2026 · Anthropic

"Largest known distillation attack": Anthropic accuses Alibaba-linked entities of siphoning Claude

In a letter to US senators, Anthropic says a 44-day campaign generated 28.8M+ interactions via ~25,000 fake accounts to illicitly extract Claude's high-value capabilities — and is calling for tougher controls on Chinese AI firms.

days of activity
(Apr 22 – Jun 5, 2026)

28.8M+

interactions with Claude

~25K

fake accounts involved

Nearly double the prior cases

Interactions with Claude (millions) — same scale, side by side.

16M+

Earlier cases combined
(3 Chinese firms, as of Feb 2026)

28.8M+

This case
(Alibaba-linked, Apr–Jun 2026)

How a distillation attack works

Flood Claude with millions of queries via fake accounts

→

Harvest the responses as training data

→

Cheaply replicate Claude's abilities in a rival model

Targeted capabilities: agentic reasoning · software engineering · long-horizon tasks

Anthropic frames it as turning "tens of billions of dollars in US investment and R&D into a massive subsidy for a geopolitical rival."

The company is asking the US government for tougher responses, including stronger export controls. The matter remains at the accusation stage — no official rebuttal has been confirmed from Alibaba or its Qwen AI lab.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.