Anthropic Accuses Alibaba of 25,000-Account Claude Distillation Attack as China Deepens AI Isolation
Updated
Updated · The Atlantic · Jul 19
Anthropic Accuses Alibaba of 25,000-Account Claude Distillation Attack as China Deepens AI Isolation
3 articles · Updated · The Atlantic · Jul 19
Summary
Anthropic told U.S. senators that Alibaba used about 25,000 fraudulent accounts to extract Claude’s “most valuable capabilities” in what it called the largest known distillation attack.
The allegation points to Chinese AI firms still leaning on stronger U.S. models even as Beijing pushes them toward domestic alternatives and limits access to foreign technology.
Beijing has blocked broad use of Nvidia’s H200 chips and is steering companies toward local semiconductors, a policy meant to build self-sufficiency but one that leaves Chinese AI developers with weaker hardware.
China’s broader strategy has produced gains in areas like EVs and open-source AI adoption, yet tighter controls on foreign partnerships and model access risk slowing innovation and widening the gap with U.S. rivals.
Can China's state-led tech fortress out-innovate the West's open ecosystem in the race for true AI supremacy?
By rejecting top US chips, is China building a self-reliant tech empire or trapping itself in second place?
"28.8 Million Outputs Stolen: Anthropic Accuses Alibaba of Industrial-Scale AI Model Theft and the Global Race to Protect Innovation"
Overview
In June 2026, San Francisco-based AI lab Anthropic accused Chinese tech giant Alibaba of large-scale AI knowledge theft, claiming Alibaba employees illicitly generated nearly 29 million outputs from Anthropic’s advanced Claude model. Anthropic described the method as 'distillation,' where a powerful AI model is used without permission to train a rival system, allowing competitors to bypass years of research and development. This accusation highlights growing tensions over the unauthorized use of proprietary AI for competitive advantage, underscoring the urgent need for stronger protections and collaboration across the AI industry, policymakers, and cloud providers.