Anthropic is calling out Alibaba for what it describes as the biggest coordinated attempt yet to steal a US AI model's capabilities through fake accounts.
In a letter shared with US senators and White House officials — a copy of which was later obtained by Bloomberg — Anthropic alleged that operators linked to Alibaba's Qwen AI lab opened nearly 25,000 fraudulent accounts and used them to systematically query Claude between April and June 2026. The goal, Anthropic said, was distillation: flood a capable model with queries, harvest the outputs, and use them to train a cheaper competitor. Anthropic did not publicly identify which Qwen models it believes benefited, nor detail how it traced the accounts back to Alibaba.
Distillation sits in a legal and ethical gray zone — somewhere between aggressive benchmarking and outright intellectual-property theft — and no court has yet drawn a clear line. By taking the accusation to Congress and the White House rather than straight to court, Anthropic is framing this as a national-security matter, not just a terms-of-service violation, which sets up a very different kind of fight.
Alibaba has not publicly responded. Qwen models have drawn consistent attention for outperforming benchmarks at lower cost — a track record that now carries a question mark.