Anthropic alleges five distillation campaigns, including Alibaba and Moonshot

Anthropic released a report Thursday alleging persistent distillation attacks by China-based AI companies, saying it observed nearly 200 million exchanges tied to five campaigns. The company said the campaigns targeted Claude capabilities including agentic tool use, coding, data analysis and logical reasoning, and that attackers used techniques to trick the model into revealing chain-of-thought traces it normally summarizes. Anthropic attributed the largest campaign to Alibaba, saying it observed 151 million exchanges between May and July 2026 across 3,500 accounts, peaking near three million per day, and linked it to training material for Alibaba's Qwen models. A separate Moonshot AI campaign, Anthropic said, routed requests through 5,000 accounts over a 10-day period, primarily targeting Opus.