Tech news in 3 minutes
Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek
Anthropic reported persistent distillation attacks by China-based AI companies, detailing nearly 200 million exchanges linked to five campaigns targeting its Claude model, in an SEO-focused news summary for readers tracking AI chip, data center, and technology business developments. The core news: Anthropic’s Thursday report alleges escalating model theft attempts, with Alibaba and Moonshot AI specifically named, as competition intensifies. The campaigns aimed to extract Claude’s chain-of-thought reasoning—including agentic capabilities, coding, and logical reasoning—by tricking the model into revealing internal thinking traces. Anthropic observed 151 million exchanges between May and July attributed
View original article
Timeline