Amodei set a China condition on the slowdown he asked the industry to accept

The pacing proposal carries a condition, and the condition is about someone else's conduct. "If we greatly restrain our AI capabilities in the belief that China will do the same, and then China defects," Amodei warned, "AI could be so powerful that such a defection could lead to their geopolitical dominance" ^[6]. As a control, that is a brake whose release depends on a running judgement about a rival that no company can audit from outside government. The Guardian's account of the call does not list its commitments or its signatories ^[20].

The same sentence runs in both directions. Trump used the race to rule out any slowdown at all. "We're leading China in AI," he said, and "We're the most sophisticated country in the world, and frankly, I want to keep it that way, because whoever wins AI, wins" ^[2][3]. Amodei used it to set the floor under a slowdown he was asking his own industry to accept ^[5].

Few have been more vocal about the threat of Chinese state-controlled AI than Amodei. He has argued for years that he pushes the technology forward so it lands with democratic governments and not oppressive regimes such as China ^[22]. Altman and Zuckerberg frame the competition in corporate or economic terms, Nvidia's Jensen Huang has encouraged the use of Chinese AI models, and Demis Hassabis of Google DeepMind has said comparatively little ^[8][9][10]. "Sam Altman doesn't have especially strong views when it comes to China. Anthropic is a totally different story," said Kyle Chan, a fellow at the Brookings Institution ^[11]. Of the five leaders the Guardian places on the question, one is described as framing it as authoritarian AI against democratic AI ^[19].

The fear he states is specific: an ultra-advanced AI that gives an authoritarian state a "swarm of millions or billions of fully automated armed drones" ^[18]. He believes such a swarm could defeat any military in the world and suppress dissent by following every citizen. "There are a number of possible arguments against the severity of these threats," he wrote in January in an essay called The Adolescence of Technology, "and I wish I believed them, because AI-enabled authoritarianism terrifies me" ^[17].

Some of this could be commercial interest. The Guardian allows both readings, writing that the industry's China concerns "may come with self-serving economic incentives, but they also reflect a strain of hawkish foreign policy toward Beijing that has become entrenched in tech" ^[13]. For planning purposes the motive matters less than the durability, and the framing has held across camps that agree on little else, joining AI safety advocates, techno-militarists and neoconservatives ^[14]. The firms have staffed for it, hiring former top US government officials from both the Biden and Trump administrations into in-house policy teams ^[15]. Anthropic and OpenAI did not respond to the Guardian's requests for comment ^[7].

This week's record covers the disposition of the president and the framing the industry has chosen. On that evidence, the pace of frontier releases is set by the companies that build them, and a board that wants a dated commitment on evaluations or release gates will be writing that date itself.