Anthropic’s Pacing Essay and the China Response: When a Safety Proposal Is Also a Containment Proposal

Dario Amodei’s three-tier pacing framework has now received verdicts at every level — and the week’s evidence demonstrates its own thesis.

What Happened

Dario Amodei’s pacing essay asked frontier laboratories to slow the rate of capability improvement and asked the United States government to widen the gap between American and Chinese AI capability through export controls and anti-distillation measures. Those are two distinct requests inside a single document. On Monday, the document received its response at the international tier — and it arrived in two registers that must not be collapsed into one.

The official response came from Foreign Ministry spokesperson Guo Jiakun at the regular Beijing briefing on Monday, September 14. The language was measured, multilateral, and deliberately unspecific: “the development of artificial intelligence bears on the common well-being of all humanity,” countries should promote AI that is “open, inclusive, universally beneficial and ethical,” and the ministry criticised the remarks as ones that “only hinder the progress of global AI governance.” That is a procedural objection framed in the vocabulary of multilateral governance — not an editorial.

The second layer is categorically different. A commentary in the Global Times — state-backed media published under People’s Daily, frequently read as a signalling instrument, but not government policy — went considerably further. As reported by GeopolitEchs, the editorial’s stated argument is that the proposal’s true agenda is “to 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.” The Cold War framing is the editorial’s own: it writes that “this silent AI Cold War is hypocritical and short-sighted”, and refers to a Cold War script. State media separately described the remarks as “hostile and baseless claims driven by commercial interests.” These are editorial registers. The ministry’s language and the Global Times’ language are not interchangeable, and attributing either to “China” without the distinction erases exactly the information that matters.

The Structural Read

The reflex when state-backed media publishes sharp language about an American executive’s policy proposals is to file it as predictable messaging and move on. That would be a mistake here — not because the editorial deserves to be endorsed, but because the essay supplies the editorial’s evidence.

Amodei writes that “pacing within democracies will be limited by the lead that US companies have over authoritarian regimes, chiefly the Chinese Communist Party” — the phrase “Chinese Communist Party” and “CCP-associated projects” are his, not a paraphrase. He writes that “if we slow down by more than this amount, then (unpaced) CCP-associated projects will pull ahead, creating significant national security risk.” The same document urges Washington to “not sell powerful AI chips or semiconductor manufacturing equipment to China, and crack down on chip smuggling operations and remote access to data centers outside China. Chips will be the main determinant of China’s AI strength.” It also asks the government to “crack down on unauthorized distillation by companies in authoritarian countries.” On the defection problem, he writes: “If we greatly restrain our AI capabilities in the belief that China will do the same, and then China defects, AI could be so powerful that such a defection could lead to their geopolitical dominance.” He told CBS News that the toughest dilemma in his proposal was exactly this — if China and other adversarial nations chose not to slow down too.

None of this is improper. Export controls are existing American policy. Chip smuggling and unauthorised distillation are genuine enforcement questions. A chief executive publicly arguing that his government should maintain a technological lead is doing something entirely ordinary, and doing it in the open. The structural observation is narrower and does not require any allegation about motive.

The global tier of the plan — international coordination resting on either ironclad verifiability or limits narrow enough that defection would not be militarily existential — required the participation of the party the same document proposes to constrain. Amodei resolved the defection risk not by treating Chinese non-participation as a condition to be negotiated, but as a risk to be hedged with export controls. That is a coherent position on its own terms. It also means “pacing within democracies” is a bloc-level agreement with a containment perimeter around it — not a global regime. The party outside that perimeter has now said publicly, at two distinct registers of explicitness, how it reads the arrangement.

That passage is the hinge of the entire framework. Amodei acknowledges the trap himself — he told CBS it was the toughest dilemma in his proposal — and then resolves it by treating non-participation as a foregone condition to be managed rather than overcome. Which means the global tier, as designed, was never a symmetric arrangement. The week’s responses — from both the ministry and the Global Times — are less a failure of the argument than a demonstration of its own thesis: some forms of coordination are legally challenging and will require government support, and every tier of the plan has now supplied evidence for that sentence.

Verdicts by Tier

Three Implications