Anthropic's chief executive called for a global slowdown on the most powerful AI models and named China as a partner; Beijing's officials, state media and engineers all said no within days.
Dario Amodei, chief executive of Anthropic, proposed in a blog post that the global AI industry slow work on its most powerful models, buying a few years to make them safer.
In the same post he named China, and proposed cutting off sales of advanced AI chips, cracking down on smuggling and restricting remote access to data centres outside the country to hold China's AI progress back.
Sam Altman of OpenAI and Elon Musk of xAI gave the idea unusual support. Mark Zuckerberg and President Trump rejected it, arguing for full speed ahead.
Beijing's answer was no
State-backed Global Times ran an editorial the next day calling it a Cold War playbook from the American tech right. Foreign Ministry spokesperson Guo Jiakun called the proposal fear-mongering and confrontation.
Yet China's own list of AI worries is growing. The third edition of its safety governance framework, released on September 14, warns that AI could optimise itself and undergo recursive self-improvement faster than humans can control.
The engineers tell a different story
The more interesting voices came from people building the models. Tao He, head of R&D at MiniMax, wrote: yesterday the mission was AGI, today we are told to slow down. Why do you always get to define the game and change the rules?
DeepSeek engineer Liu Shengyu fears a different ending: a handful of companies holding the most powerful AI while everyone else falls further behind. He wants powerful AI cheap and widely available, and does not trust Anthropic or OpenAI to deliver that.
Huawei rotating chairman Eric Xu put it more bluntly: American labs may be meeting risks Chinese developers cannot yet perceive, so China's job is not to slow down but to catch up.
In one line: the people urging everyone else to brake are usually the ones sitting furthest forward.
Why it matters
Slowing down AI sounds like a shared human interest, but it first has to answer who sets the pace, by whose standard, and who actually complies. This exchange lays that layer bare: safety rhetoric and a lead in the field are hard to prise apart in practice.
Curated from high-quality sources, with concise summaries and key takeaways.
