Anthropic has accused Chinese tech giant Alibaba of running what it calls the largest-known "distillation" campaign against its Claude AI models, according to reporting from Reuters, the BBC, and others.
In a letter to US senators, Anthropic alleged that operators linked to Alibaba's Qwen AI unit used nearly 25,000 fraudulent accounts to generate roughly 28.8 million interactions with Claude, according to Firstpost and Crypto Briefing. Nikkei Asia reported Anthropic framed the effort as the "largest known distillation attack" on its models.
Distillation is a technique where one AI model is effectively trained on the outputs of another. As TechRadar put it, Anthropic believes Alibaba may have copied Claude's capabilities "just by asking questions" — feeding it millions of prompts and using the responses to improve a rival system. Anthropic describes this as illicitly extracting its model's capabilities.
The accusation lands amid wider tension over AI and national security. According to India Today and MSN, the disclosure came as the US restricts Chinese access to advanced models including Mythos and Fable 5, sharpening concerns about how cutting-edge AI flows across borders.
Markets reacted: Investor's Business Daily reported that Alibaba's stock slid on the news. As of the reporting, Alibaba had not publicly responded to the allegations, according to CNBC-TV18.
It is worth noting these are allegations from Anthropic, a direct competitor, and not findings from an independent investigation or court.
Why it matters: if a leading AI lab can have its work siphoned off at this scale through ordinary-looking queries, it raises hard questions about how AI companies protect their most valuable products — and feeds a growing US-China fight over who controls frontier artificial intelligence.