Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek

Anthropic has released a report detailing persistent and escalating "distillation attacks" by Chinese AI companies, including Alibaba, Moonshot AI, and DeepSeek. The attacks, totaling nearly 200 million exchanges, aim to extract the internal reasoning (chain of thought) of Anthropic's Claude model to train competing models. Alibaba's campaign was the largest, with 151 million exchanges attributed to it, targeting Claude's agentic, coding, and reasoning capabilities. Moonshot AI's campaign appeared to route requests from Chinese military entities. The attackers used sophisticated techniques to bypass Anthropic's defenses, which typically hide the model's raw thinking process.

(Source：TechCrunch)