Anthropic has accused Alibaba of running the largest campaign ever to illicitly extract Claude’s capabilities, and it’s urging Congress to do something about it.
Ars Technica obtained a June 10 letter to Senators Tim Scott and Elizabeth Warren shared ahead of a Senate committee hearing on “AI and the American Dream.” In it, Anthropic presented what it called new and confidential evidence of the biggest distillation campaign it has ever measured.
Between April 22 and June 5, operators affiliated with Alibaba and its Qwen AI lab generated more than 28.8 million exchanges with Claude using nearly 25,000 fraudulent accounts. The campaign targeted some of Claude’s most valuable capabilities: agentic reasoning, software engineering, and long-horizon planning.
The attack followed Anthropic’s release of Mythos, a cybersecurity AI model that was subsequently restricted in foreign markets. Anthropic’s letter frames the broader race to match US AI capabilities as a national security concern.
This isn’t Anthropic’s first accusation against Chinese AI firms. Back in February, the company claimed DeepSeek, Moonshot AI, and MiniMax collectively made over 16 million Claude exchanges through roughly 24,000 fake accounts. Critics argued that AI companies themselves rely on similar techniques when training models, but Anthropic maintains there’s a line between legitimate distillation and unauthorized extraction via fraudulent access.
