'Fearmongering' Meets 'Cold War Playbook': Beijing Formally Rejects the US AI Slowdown

Two days after Anthropic CEO Dario Amodei published a 3,800-word essay urging the world’s AI labs to deliberately slow down, Beijing has delivered its answer — and it is a hard no.

On Monday, September 14, China’s Ministry of Foreign Affairs dismissed the growing chorus of US tech executives calling for a pause in frontier AI development, describing their warnings as “fearmongering” that serves nobody’s interest. The statement, issued by spokesperson Guo Jiakun at a regular press conference, marks the first formal state-level response to the pacing push that has dominated the AI news cycle since Amodei’s essay landed on Saturday.

“Fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance and serve the interests of no one,” Guo said, according to multiple wire reports. He urged all parties instead to “promote an open, inclusive and beneficial-to-all” development of artificial intelligence.

What triggered it

The immediate target of Beijing’s ire is Amodei’s Saturday essay, “We Must Pace the Frontier,” in which the Anthropic chief executive argued that AI companies must “slow the pace at which we improve the capabilities of AI models” to give safety work time to catch up. The essay laid out a concrete plan: third-party evaluations of frontier systems before deployment, tighter controls on compute, and — critically, from Beijing’s perspective — export restrictions.

The most contentious passage was unambiguous about China. Amodei wrote that any slowdown by “democratically governed countries” should be calibrated to preserve the lead US AI firms currently hold over Chinese competitors. His prescription included a strict ban on selling powerful AI chips or semiconductor manufacturing equipment to China, a crackdown on chip smuggling and remote access to data centers, and what he framed as “cooperation with China” only in narrow, safety-specific channels.

Global Times, the state-run tabloid that often carries Beijing’s sharpest messaging, went further than the Foreign Ministry. In an editorial published over the weekend, it said the essay was “packed with containment provisions targeting China and is, in essence, a ‘Cold War playbook’ for the AI sector.” A follow-up piece counted that Amodei mentioned China 12 times in the essay, concluding that the “true agenda” was “to attempt to curb China’s AI development through technological barriers and regulatory monopolies.”

Two narratives collide

The exchange crystallizes the core dilemma of the entire slowdown debate — and both sides know it.

For Amodei and the executives who echoed him, the China question is the plan’s “toughest dilemma,” as he acknowledged in a follow-up interview. If US labs unilaterally pause and Chinese labs don’t, the pause accomplishes nothing except handing over the frontier. That is why his essay pairs the slowdown call with containment measures: slow down, but slow down with a moat.

Beijing’s counterargument inverts the framing entirely. In its telling, the “safety” language is the moat. Chinese officials and state media have consistently argued that US export controls, investment restrictions, and now pacing proposals are not about safety at all — they are about using regulatory tools to lock in a technological lead that fair competition might otherwise erode. The speed with which Chinese labs have closed the efficiency gap — training competitive models on constrained compute budgets — is Exhibit A in that argument.

There is also a domestic-audience dimension. The pacing push arrives amid a bruising week for AI equities in Asia, a US presidential statement rejecting any slowdown, and growing public anxiety stoked by weeks of high-profile AI warnings. A unified, defiant response plays well at home and reinforces China’s preferred venue for AI governance: multilateral, UN-centric processes where it has more diplomatic traction — such as the recent World AI Conference and the growing list of signatories to its AI treaty initiative — rather than bilateral deals negotiated from a position of claimed US superiority.

Why the timing matters

The statement lands days before a planned mid-September US–China dialogue on frontier AI safety risks, which Reuters first reported on September 4. Those talks — a follow-up to earlier Geneva-track discussions — are supposed to establish shared understandings on risk management for frontier models, with sources indicating the issue could also surface at a potential Trump–Xi meeting later this month.

A public spat over whether the entire premise of “pacing” is good-faith safety advice or disguised containment does not make those talks easier. US negotiators arrive seeking Chinese buy-in for exactly the kind of measures Amodei outlined: transparency about frontier capabilities, incident notification, and mutual restraint. Beijing has now, on the record, pre-rejected the intellectual framework behind all three.

Yet the talks have not collapsed before they began. Notably, Guo’s statement — while blunt — stuck to the Foreign Ministry’s standard governance vocabulary (“open, inclusive, balanced, beneficial to all”) rather than announcing any withdrawal from the dialogue. State media criticism of Amodei personally is also narrower than it might appear: Beijing has an interest in keeping the channel open, not least because Chinese labs face the same loss-of-control risks Amodei warns about, and incidents on either side of the Pacific don’t respect firewalls.

The uncomfortable middle

For everyone caught between the two positions, the week’s escalation narrows the space for a genuine safety agenda.

The labs that genuinely want to slow down now face a worse bargain: pause unilaterally and cede the frontier, or accept Beijing’s framing that any safety-motivated restraint is suspect. Regulators who want international coordination on frontier risks — the UK’s proposed ASI moratorium, the EU’s implementation of the AI Act, UNESCO’s ethics forum — must now navigate a geopolitical framing war in which “safety” itself has become a contested term. And researchers watching agent incidents multiply get the worst of both worlds: the risks are real, but the loudest risk-messaging is now entangled with industrial policy.

The irony is sharp. Amodei’s essay argued that a slowdown is only credible if paired with containment, because trust between the two AI superpowers is absent. Beijing’s reply demonstrates exactly how absent it is — and in doing so, makes the multilateral governance process it claims to champion the only path left standing.

Whether that process can bear the weight is the question the mid-September talks will begin to answer.