Anthropic announces large-scale efforts to acquire Claude
The article discusses the "Anthropic" company's campaign to uncover theft of artificial intelligence models' capabilities, particularly by companies like "Alibaba" and "Moonshot." Anthropic revealed extensive activities involving around 200 million interactions aimed at distilling their advanced models to extract abilities such as reasoning and data analysis through attacks that involve submitting disguised requests and deceiving the models. The most significant campaign activities were conducted by Alibaba between May and July 2026, with approximately 151 million interactions, while Moonshot utilized thousands of accounts to analyze CCTV recordings and other models. These campaigns highlight the escalating global competition to steal and develop AI model capabilities and underscore the need for security measures to combat such advanced extraction operations.
Notice: This Is an AI-Generated Summary
Comments (0)