Anthropic chief executive Dario Amodei has called for frontier AI companies to slow the rate at which they improve model capabilities.[1] His proposal is not a simple halt: it combines embedded independent evaluators, coordination between companies in democratic countries and, eventually, international agreements on the most dangerous capabilities.[1]
Amodei also recognises the central geopolitical problem.[1] A unilateral slowdown could sacrifice the United States’ lead if China continued advancing, while any wider agreement would need unusually strong verification.[1] His plan therefore pairs pacing with tighter controls on advanced chips, model theft and unauthorised distillation.[1]
President Donald Trump has rejected the slowdown argument.[2] He told reporters that the United States was leading China and wanted to keep that position because “whoever wins AI wins”.[2][5] That stance is consistent with the White House AI Action Plan, which describes AI leadership as an economic and national-security contest and calls for faster innovation alongside monitoring of emerging risks.[3]
China has rejected the US executives’ framing too, but for a different reason.[4] Its Foreign Ministry said “fear-mongering, confrontation and vicious competition” would obstruct sound global AI governance, while calling for open and inclusive development.[4] Chinese state media and researchers have separately argued that a slowdown could preserve the advantage of established US companies rather than create a neutral safety regime.[5]
The disagreement exposes the weakness in any voluntary pause. A frontier laboratory can slow its own work, but it cannot guarantee that competitors or governments will follow.
From the strategic perspective raised by Trump, restraint without enforceable international participation risks giving rivals time to close the gap. From the laboratories’ perspective, continuing at full speed could let capabilities outrun safety work. Both concerns can be true at the same time.
Supporting sources:
[1] Dario Amodei: We Must Pace the Frontier
[2] Associated Press: Trump dismisses idea of new AI guardrails
[3] The White House: America’s AI Action Plan
[4] Chinese Foreign Ministry press conference, 14 September 2026
[5] NBC News: China rejects AI slowdown calls
Source: Dario Amodei