Anthropic has urged the US to widen its edge over China in artificial intelligence capability to “avoid authoritarian AI leadership”, a warning which is being criticised by some industry experts as “irresponsible” and self-serving.
In a blog post published on Thursday as the first day of the high-stakes summit between Chinese President Xi Jinping and US President Donald Trump wrapped up in Beijing, the American AI firm urged the US and its allies to tighten export controls and curb the so-called distillation attacks originating in China – a technique where smaller models are trained using the outputs of larger, more advanced systems.
Anthropic said that locking in a 12-to-24-month lead in frontier capabilities by 2028 would be “enormously advantageous”. This gap would allow the US to increase its engagement with AI experts in China in areas such as AI safety and governance, according to Anthropic.
The prospect of the Communist Party leading in frontier AI is “among the greatest threats” to the future of humanity, “enabling repression at a scale that humans alone could not achieve”, Anthropic wrote.
Anthropic, developer of the cutting-edge Claude models, is known to be a vocal critic of China’s AI advancement and a strong advocate for US chip export control measures.
While the company’s latest remarks raise legitimate concerns about misuse that merits serious engagement, its “arms-race framing pushes us in the wrong direction at exactly the wrong moment”, said Alvin Wang Graylin, Digital Fellow at the Stanford Institute for Human-centered Artificial Intelligence and senior fellow at the Asia Society Policy Institute.