China Isn’t Buying Silicon Valley’s Call for an AI Slowdown

Dario Amodei published We Must Pace the Frontier. The Anthropic CEO’s blog post argues for a global slowdown in AI development—but with one critical caveat. Any coordinated pause, he writes, must ensure the US and its allies maintain "as large as possible" a lead over autocracies like China. That single clause was met with skepticism in Beijing. It reinforced Chinese officials’ and researchers’ skepticism about the motives behind Silicon Valley’s safety concerns.

This isn’t about whether AI is dangerous. It’s about who gets to define safety—and whether a slowdown can ever be neutral when AI is the ultimate strategic asset.

Amodei’s post isn’t subtle. Amodei wrote that a key part of the slowdown strategy should be ensuring the US and other democracies maintain ‘as large as possible’ a lead in AI development over autocracies like China. That’s not a throwaway remark. It’s the entire framing. The slowdown isn’t presented as a universal good but as a conditional advantage—one that preserves American dominance while asking others to hit the brakes.

Compare this to past AI safety proposals. No geopolitical strings attached.

Amodei’s post breaks from this tradition. The mechanism for the slowdown? Unclear. The metrics for success? Vague. The only concrete detail is the insistence on maintaining a US lead. That’s not an oversight. It’s the point.

Chinese officials aren’t rejecting the idea of AI risks. They’re dismissing the motives behind Amodei’s proposal. Alvin Wang Graylin, a prominent Chinese AI researcher, put it bluntly: "Amodei’s essay makes the AI pacing conditional on increasing the US lead over China, which feels inauthentic to Chinese readers. " That word—inauthentic—is telling. It’s not that China thinks AI is safe. It’s that they see this as a play for dominance, not a genuine safety measure.

This skepticism runs deep. Michelle Nie, a researcher quoted in Wired, notes that Beijing doesn’t dismiss AI risks outright, but it does reject the narrative that AGI poses existential threats. The CCP’s primary concern? The threat AI poses to its own rule. These are the types of risks that concern Chinese policymakers, not rogue AGI scenarios.

Why?

The US and China aren’t just disagreeing on how to handle AI risks. They’re operating from entirely different risk models.

In Silicon Valley, the discourse is dominated by existential threats. Rogue AGI. Human extinction. Alignment problems that could spiral out of control.

China’s priorities are different. Michelle Nie’s observation cuts to the core: "The risk Beijing seems most worried about is the threat AI could pose to CCP rule. " That’s not a fringe view. It’s baked into China’s regulatory approach. The concern isn’t that AI will become sentient and turn against humanity. It’s that AI will pose risks to political stability.

This divergence shapes everything. One side fears a misaligned superintelligence. The other fears a misaligned populace.

Nie’s analysis hits the mark: "Chinese policymakers tend to read US companies’ calls for AI safety as instrumental. " That’s a polite way of saying they see it as a cover for competitive advantage.

This isn’t paranoia. It’s historical precedent.

The call for a slowdown isn’t just about safety—it’s about locking in a US lead. That framing reinforces Chinese skepticism about US motives.

The irony? Amodei’s explicit geopolitical condition doesn’t challenge this perception. It confirms it. But by tying the proposal to a US lead, Anthropic reinforced Chinese skepticism about US motives.

Amodei’s proposal didn’t have to include a geopolitical condition.

But Amodei didn’t do that. He explicitly tied safety to a US lead. That wasn’t an accident. It was a strategic choice—one that reinforces Chinese skepticism.

Why?

The current state of play reinforces this dynamic. The US leads in frontier AI development. It’s unthinkable.

A global AI slowdown requires coordination. Coordination requires trust. And trust is in short supply between the US and China.

China’s approach to AI safety reflects its risk model. These aren’t just technical safeguards. They’re tools for political stability.

The US approach is different. Export controls on advanced chips (e.g. These measures aren’t about existential safety. They’re about maintaining a strategic edge.

Neither side can accept the other’s framework without sacrificing its core interests. Safety, in this context, isn’t a shared goal. It’s a weapon.

Anthropic’s proposal faces deep skepticism in China. That’s not a bug—it’s a feature. The explicit geopolitical framing reinforces Chinese skepticism about US motives.

It’s widening.

Or do they adopt China’s approach?

These don’t require geopolitical alignment. They’re just good engineering practices. But even here, the trust deficit looms large.

Amodei’s proposal faces deep skepticism in China.

It’s Washington. It’s Brussels. It’s Tokyo. By framing the slowdown as a way to preserve a US lead, Amodei isn’t just making a safety argument. He’s making a strategic one.

The real question isn’t whether a global AI slowdown is possible. It’s whether safety can ever be disentangled from geopolitics. The answer, it seems, is no. AI governance isn’t just about alignment or interpretability. It’s about power. And in a world where AI is the ultimate strategic asset, every safety call is also a geopolitical play.

That’s the lesson from Amodei’s post. The slowdown call wasn’t doomed by technical challenges or scientific disagreements. It was doomed by the reality that, in a zero-sum world, safety is just another front in the tech cold war. The question now is whether anyone can change that—or if the divide is already too deep.