Updated

Anthropic said it detected unauthorized efforts by China-based AI labs, including Alibaba and Moonshot AI, to use millions of exchanges with its Claude models to help train and improve their own competing systems. The disclosure adds to a broader debate among U.S. AI companies over model 'distillation,' in which rivals extract knowledge from frontier models to build cheaper competitors, a practice several AI giants have accused Chinese firms of engaging in. At Y Combinator's Demo Day, CEO Garry Tan pushed back against alarm over distillation, arguing companies should largely 'do nothing' about it rather than treat it as an existential threat. The comments highlight a split in Silicon Valley over how seriously to treat both the competitive and security risks posed by China's rapidly advancing AI sector.