Why Slowing China's AI Could Backfire on Everyone

A quiet debate is happening among the people building the world's most powerful AI. Dario Amodei, who runs Anthropic — maker of the Claude chatbot — wrote that slowing China's AI progress might actually help international cooperation. It sounds backwards. Wouldn't frustrating a rival make them trust you less? His logic: if America keeps a clear lead, China has a reason to sit down and negotiate rules. If the two are neck and neck, neither has much reason to hold back.

The essay then digs into why smart people disagree so wildly about AI danger. One reason is they want different things. The other is they want the same thing but their gut sense of the odds is skewed. Incentives quietly bend those odds. If your company is about to sell shares to the public, or you invest in AI startups, you may unconsciously decide the risk is smaller than it is. Nobody thinks they're doing this. That's exactly why it works.

History suggests countries aren't immune either, even though having many voices makes them more rational than any one person. In 1941, Japan attacked Pearl Harbor knowing a long war with America was probably unwinnable — backing down after an oil embargo felt like national humiliation. Austria-Hungary turned a single assassination into a world war in 1914. America turned 9/11 into the Iraq War in 2003. Pride and hurt feelings drove all three.

So if the US tries to slow China's AI, it might gain negotiating leverage. Or it might just create a resentful rival. The honest answer is nobody knows. What is clear is that the rules being written right now will decide how AI touches your job, your privacy, and your safety — and they're being shaped by humans with blind spots, not machines.