Can AI safety pacing work without government cooperation?

Alberto Romero's essay on The Algorithmic Bridge argues that Dario Amodei's "We Must Pace the Frontier" proposal — to slow AI capability advances while safety catches up — ran immediately into a political wall. Within days of publication, "President Donald Trump, who notably holds one key to this solution—the other key holder being Xi Jinping—said no to the slowdown," declaring that "whoever wins AI wins." Xi Jinping, for his part, "called Amodei's call to a slowdown a 'Cold War playbook,'" and the Global Times framed it as an attempt "to curb China's AI development through technological barriers and regulatory monopolies, uphold Washington's monopolistic hegemony in cutting-edge technology, and exclude China from the global AI governance system."

Romero's reasoning is that Amodei's three-stage plan (embedded evaluators, Western-lab coordination, eventual coordination including China) depends on state cooperation that national leaders have little incentive to grant once AI becomes a tool of geopolitical competition. He casts the standoff as "Oppenheimer vs Truman" — technologists' safety concerns meeting state ambition — and concludes "technology is power but not more so than power itself." On that basis he predicts the pacing plan "ain't happening," not because its technical premises are wrong but because the political actors who would need to ratify it have already said no.

This sits downstream of Should AI capabilities growth be deliberately slowed to allow safety work?, supplying the proposal's reception rather than its content: where that note states what Amodei asked for, Romero's essay tracks what happened once heads of state weighed in. It also runs parallel to Can industry self-regulation slow AI without government enforcement?: Karpf's objection is that embedded monitoring is toothless without government force, and Romero's account shows the two governments that could supply that force — the US and China — both declining within days of publication.

The excerpt is commentary, not reporting with independent verification — Romero does not quote primary transcripts of Trump's or Xi's remarks beyond the lines given, and he states plainly that he takes Amodei's intentions "at face value" rather than establishing them independently. It also doesn't establish that the plan is permanently dead — only that its first political test, immediately following publication, returned two rejections. The implication Romero draws, cautiously, is that voluntary industry pacing proposals face a ceiling set by state competition for AI advantage, not by remaining technical disagreement.