Attacked from two fronts: Amodei's call to slow AI meets resistance in both Beijing and Washington

Attacked from two fronts: Amodei's call to slow AI meets resistance in both Beijing and Washington

Dario Amodei's call to slow the development of frontier AI models was published on 12 September. Three days later, it had become a matter of great-power politics: China's state-backed Global Times calls the essay a "Cold War playbook," the foreign ministry dismisses the warnings as "fearmongering," and Trump attacks them from the right as a conspiracy that "benefits China." All of this happens days before the Trump–Xi meeting on 24 September.

What happened

Anthropic CEO Dario Amodei published the essay "We Must Pace the Frontier" on 12 September, whose core argument is that AI capability progress must be deliberately slowed: "We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain" (Amodei, primary source). According to The Straits Times, the essay was later endorsed by OpenAI CEO Sam Altman and Elon Musk (5e735f9c).

The reaction came quickly, and from two sources that rarely cite the same concern:

• Global Times, a state-backed Chinese outlet, wrote in a 14 September editorial that the essay "may look like a rational statement focused on safety risks, but it is really a 'Cold War playbook' targeting China." The editorial calls it a "silent AI Cold War" that is "hypocritical and short-sighted," and claims that excluding China from global AI innovation would "significantly increase the trial-and-error costs and risks of loss of control in global AI development" (via The Straits Times, [5e735f9c]). The editorial further claims that Amodei's real agenda is "to attempt to curb China's AI development through technological barriers and regulatory monopolies." Note: all quotes from the editorial come via secondary sources — The Straits Times and NBC News; the editorial itself is not available here in a first-hand source.
• The foreign ministry followed up. Spokesperson Guo Jiakun said at a regular Monday briefing: "Fearmongering, confrontation and malicious competition will only disrupt the process of global AI governance and serve no one's interests" (via NBC News, 29449aec).
• Trump attacked Amodei directly on Truth Social on Monday, according to the New York Post: "There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS!" (5ef6894b).

What Amodei actually proposes

In the essay, Amodei argues that risk management cannot keep up with capability development, and that progress must therefore be "paced" so that prevention has time to catch up ([cf36150b]). According to NBC News, he also writes that a "Chinese lead in AI would pose grave danger for the United States and the world," and urges the US to restrict sales of key technology to China ([29449aec]).

This is where the pacing argument shifts from security policy to power politics: the essay explicitly links tempo control to export controls aimed at China. One clarification is warranted — Amodei's detailed recommendations on chip export controls and measures against so-called model distillation are known only through secondary summaries from NBC and The Straits Times; the available essay excerpt concerns the pacing argument itself, not these sections.

What complicates the picture: China is not rejecting security thinking

The simple narrative — that China rejects AI risk because it is politically convenient — sits poorly with China's own planning. Since September 2024, China has explicitly accounted for future loss-of-control scenarios in an AI security framework drawn up under the Cyberspace Administration of China (CAC), a framework that was expanded in September 2025 (Reuters via MSN, 4d45dd1b).

The same weekend the rejections came, China's security minister Chen Yixin said, according to NBC News, that China must "accelerate the establishment of a system for preventing and controlling AI security risks" — while, in the state-run China Cyberspace magazine, describing AI as "a new track for strategic rivalry" and writing that China "must engage extensively in international cooperation" ([29449aec]).

That does not mean Beijing sympathizes with Amodei. The point is more precise: China is not rejecting the risk itself, but the framing Amodei places around it — that security arguments should serve as justification for export controls and technological containment. Security planning on its own is preparedness; coupled with export restrictions, it becomes, in Beijing's reading, a weapon.

Converging rhetoric from opposing positions

The most striking feature of the material is how Trump and Global Times, otherwise on opposite sides of nearly everything, arrive at the same analytical conclusion: that American AI safety regulation benefits China. Trump says it as conspiracy theory — "the only one that is happy about it is China." Global Times says it as accusation — that the pacing call is a cover for hindering Chinese development.

These are documented statements, not a conclusion that they are right. But the convergence has a practical consequence: Amodei's core message — that capability progress can outgrow risk management, on both sides of the Pacific — is now politically unsafe in both capitals. In Washington it appears as hostile propaganda; in Beijing as hegemonic thinking. This leaves little political room for precisely the international security coordination that Amodei, and for that matter Chen Yixin, both invoke.

What remains uncertain

The three-way clash lands just before the Trump–Xi summit on 24 September in Washington, which NBC News describes as a meeting where the differences in AI approach "is likely to be a central issue" ([29449aec]). Several things remain unsettled:

• The content of the meeting. Nothing in the available material says anything about outcomes, and it would be wrong to assume any.
• The timing of a planned US–China AI dialogue is unclear: the sources suggest different timings ("later this month" at NBC, "mid-September" at The Straits Times), and it is also unconfirmed that the talks will take place at the stated time.
• Whether the pacing argument has a political future in the US. Amodei's position has backing from Altman and Musk, but has now been attacked by a sitting president. Since the argument is interwoven with China criticism, it is uncertain whether it can be untangled from that context.

The main picture after three days, then, is this: the question of how quickly frontier models can safely be developed is no longer a technical or security-policy debate. It has become part of the rivalry between the US and China — and that is precisely what makes it harder to resolve.