Anthropic Accuses Alibaba of Largest-Ever AI Distillation Attack: 25,000 Fake Accounts, 28.8 Million Exchanges
A letter to the U.S. Senate just blew the lid off the biggest "copy-my-homework" scandal in AI history.
Three Key Takeaways
On June 24, Anthropic sent a letter to the U.S. Senate Committee on Banking, Housing, and Urban Affairs accusing Alibaba's Qwen AI lab of orchestrating the largest known distillation attack against Claude. According to the letter, between April 22 and June 5, 2026, operators affiliated with Alibaba used roughly 25,000 fraudulent accounts to conduct 28.8 million exchanges with Claude. To bypass Claude's China access restrictions, the attackers used commercial proxy services to mask the geographic origin of their API calls.
The attack specifically targeted Claude's two most commercially valuable capabilities: agentic reasoning and software engineering. Distillation attacks work by systematically querying an advanced model and using its responses as training data to improve a cheaper competitor — essentially skipping millions of dollars in R&D by harvesting someone else's output. Anthropic believes the extracted data was used to improve Alibaba's Qwen model series. Alibaba denies any wrongdoing.
This isn't an isolated incident. Back in February 2026, Anthropic disclosed previous distillation attacks linked to other Chinese AI labs, including DeepSeek, Moonshot, and MiniMax. But Alibaba's campaign dwarfs all prior cases — 28.8 million exchanges over 45 days works out to more than 630,000 interactions per day.
WangDou's Take
25,000 fake accounts, 28.8 million exchanges, 45 days — this isn't "academic curiosity," this is an industrial-scale knowledge extraction pipeline. Do the math: each account averaged 1,152 conversations with Claude, an order of magnitude beyond any normal usage pattern. What's more telling is what they targeted — not chat, not creative writing, but agentic reasoning and coding, the two hardest and most expensive muscles to train in today's AI race. Alibaba denies it, but Anthropic delivered the letter straight to the Senate, which means this is almost certainly about to become another bullet in the U.S.-China AI rivalry narrative.
Source: CNBC · Tom's Hardware · eWeek