Anthropic Alleges Chinese AI Firms Used Fraudulent Accounts to Access US Models
Anthropic has alleged that Chinese artificial intelligence developers utilized fraudulent accounts and proxy networks to extract reasoning data from its Claude model, with some instances suggesting the US model served as a hidden backend for their applications. This comes amid broader concerns from US officials regarding Chinese technology practices.
)
According to the South China Morning Post, a central question surrounding China's technology sector involves whether Chinese AI companies have been secretly accessing US models to narrow the technology gap. This inquiry follows a series of accusations originating from Washington and the American AI laboratory Anthropic.
A joint advisory issued by the National Security Agency, the FBI, and the Cybersecurity and Infrastructure Security Agency last month indicated that Chinese AI developers have been systematically drawing out capabilities from US frontier models to lower their own expenses related to computing, electricity, and research.
Furthermore, US Treasury Secretary Scott Bessent issued a warning in July stating that Chinese companies were involved in large-scale "distillation."
Specifically, Anthropic asserted that Chinese AI developers employed fraudulent accounts and proxy networks to siphon reasoning data from its primary model, Claude. In certain instances, it was claimed that the US model was utilized as a covert backend for the developers' own applications.
)
)
)