China Pushes Back on Dario Amodei’s Call to Slow the AI Race

China is rejecting Anthropic CEO Dario Amodei’s push for the United States and Beijing to jointly slow frontier AI development, arguing that the proposal would unfairly constrain Chinese firms while preserving U.S. advantages. The dispute matters because it highlights how hard it may be to turn shared fears about advanced AI into real cooperation between the world’s two most powerful AI rivals.

Over the past week, leading Silicon Valley figures have shifted from warning about an AI arms race with China to calling for some form of international pacing or slowdown. Beijing’s response has been skeptical and blunt: Chinese officials and industry voices say they are not opposed to AI safety, but they do not trust a framework that appears designed to keep China behind.

The clash underscores a deeper problem in global AI diplomacy. Both countries say they worry about misuse, cyberattacks, and other dangers, but they differ sharply on what the greatest risks are and who should set the terms for managing them.

What triggered the latest AI diplomacy dispute?

The latest row began after Amodei published an essay arguing that frontier AI has become so powerful that the U.S. and other democracies should work with China to slow its advance. His central claim was that the two sides need coordination to reduce the chance that highly capable systems are deployed recklessly or become impossible to control.

But his proposal quickly ran into a political reality: Chinese readers and officials interpreted it not as a neutral safety pitch, but as a demand that China accept restrictions while the U.S. preserves a lead.

“Global pacing will require cooperation with China,” Amodei wrote, while framing the country as the most advanced autocracy in AI and arguing that democracies should retain the largest possible lead over it.

That combination — a call for cooperation paired with a strategy built around widening the U.S. lead — is exactly what has fueled backlash in Beijing.

Why does China see the proposal as unfair?

China’s reaction is rooted in the belief that the U.S. is not offering a genuine partnership. Instead, Chinese policymakers and researchers see a proposal that would first limit Chinese capabilities, then invite negotiation on terms set by Washington and Silicon Valley.

Kevin Xu, founder of Interconnected Capital and a former White House staffer, says Chinese AI safety thinkers share many of the same concerns as their U.S. counterparts. The divide, he argues, is over power and status. Beijing is not interested in a framework where it is expected to accept restraints before being treated as an equal.

Alvin Wang Graylin, a senior fellow at the Asia Society Policy Institute and a digital fellow at Stanford’s Institute for Human-Centered AI, offered a similar critique. He said Amodei’s essay makes the slowdown idea feel conditional on increasing America’s lead over China, which makes it difficult for Chinese audiences to view the message as sincere.

That criticism goes to the heart of the standoff. For Beijing, the issue is not whether AI is risky. The issue is whether a safety framework is being used as a geopolitical instrument.

How did Chinese officials respond publicly?

Chinese officials wasted little time in signaling their objections. At a press conference on Monday, Chinese Foreign Ministry spokesperson Guo Jiakun described the proposal as “fear-mongering, confrontation, and vicious competition,” saying it would hurt international cooperation on AI governance.

The ministry followed up the next day by reiterating that China believes AI development and security must be handled together, a phrase that reflects Beijing’s preference for balancing innovation with control rather than adopting a U.S.-style existential-risk narrative.

Those remarks were not a flat rejection of AI safety. Instead, they reflected a refusal to accept a framework that appears to make China the party that must first give ground.

What worries Beijing about AI?

Beijing’s main concerns differ from the most alarmist debates in Silicon Valley. While many U.S. frontier-lab executives talk about the possibility of catastrophic or even civilization-ending scenarios, Chinese leaders are more focused on whether AI could destabilize society or weaken state authority.

Michelle Nie, a visiting fellow at the Center for a New American Security, says Beijing is skeptical of claims that AI poses extreme existential risks to humanity. Chinese officials are beginning to acknowledge loss-of-control scenarios, she notes, but their dominant concern is whether AI could undermine Communist Party rule.

In practical terms, that means Chinese authorities are watching for:

• political misinformation
• cyberattacks and espionage
• social unrest
• military applications that could alter the balance of power

That list is notably different from the apocalyptic framing that often dominates AI safety conversations in the West. It also helps explain why Chinese officials view American calls for pacing with suspicion.

Nie said Chinese policymakers often interpret U.S. company appeals to AI safety as strategic rather than altruistic, believing they may be used to strengthen corporate valuations or slow Chinese competitors.

How are Chinese AI companies thinking about the race?

China’s AI industry has generally moved in a different direction from the closed-model, high-capital approach favored by leading U.S. labs. Many Chinese firms have leaned into lower-cost systems and open-weight releases, a strategy that broadens access and reduces dependence on a small number of gatekeepers.

That orientation is part ideology, part business model, and part response to U.S. export controls and chip restrictions. It also helps explain why warnings from Anthropic and OpenAI can land poorly in China’s AI community.

A researcher at DeepSeek published an essay over the weekend that captured this frustration. The researcher argued that AI could push society in one of two directions: broad access that raises living standards, or concentration of power in the hands of a few companies that control the most advanced systems and the infrastructure around them.

The DeepSeek researcher said the most advanced intelligence should be made cheap and available to everyone, and expressed distrust of Anthropic and OpenAI as custodians of frontier AI.

That view is important because it shows the Chinese AI debate is not just about nationalism or rivalry. It is also about who should benefit from the technology and whether safety arguments are being used to entrench market power.

Why Silicon Valley’s message may be backfiring

The irony of the current debate is that some U.S. executives are making stronger arguments for cooperation at the same time they are endorsing policies that make China feel singled out. Amodei’s essay urged the U.S. to crack down on chip smuggling and prevent Chinese companies from distilling American models without permission.

Those are substantive security concerns in Washington. In Beijing, though, they reinforce the perception that AI safety is being folded into broader competition policy.

Amodei also argued that if the U.S. weakens China’s AI sector, Beijing may later have more reason to negotiate. That argument assumes China would respond to pressure by seeking a deal. Chinese policymakers, however, may interpret it as confirmation that the U.S. intends to use safety rhetoric to preserve strategic advantage.

The result is a familiar diplomatic trap: the more one side frames cooperation as dependent on the other side accepting constraints, the less credible that cooperation becomes.

What is the core strategic mismatch?

The core strategic mismatch is that the U.S. is asking China to help manage a technology race while also trying to preserve American dominance in that race.

From Beijing’s perspective, that is not a neutral governance proposal. It is a request to accept asymmetry first and diplomacy later. That makes it hard for Chinese officials to separate safety from rivalry, even when they agree that AI deserves tighter oversight.

Can the United States and China still cooperate on AI safety?

Yes, limited cooperation is still possible, but it is likely to be narrow, practical, and heavily verified. Despite the public rhetoric, both governments have identified overlapping concerns, including cybersecurity, model misuse, and the unpredictability of increasingly advanced systems.

According to reports cited in the debate, preliminary talks are already taking place ahead of Chinese leader Xi Jinping’s planned trip to Washington later this month to meet with President Donald Trump. That suggests there is at least some appetite for discussion, even if expectations remain low.

Michelle Nie said the most realistic path forward is to focus on specific issues where both countries can make credible commitments and independently check whether those commitments are being followed.

Nie said progress may depend on building a verification mechanism that allows each side to trust the other’s compliance on limited, clearly defined AI safety measures.

That is a much narrower objective than broad global AI governance, but it may be the only one available in the near term.

What issues could be easier to negotiate?

Some areas are more likely to produce incremental progress than others. These include:

1. cybersecurity safeguards around frontier models
2. limits on certain forms of misuse
3. reporting channels for dangerous incidents
4. technical exchanges on model behavior and loss-of-control risks

These topics are less politically charged than questions about who leads in frontier AI, and they map more closely to practical threat reduction than to geopolitical positioning.

Table: the main fault lines in the U.S.-China AI debate

How does this fit into the broader AI arms race?

This debate is part of a larger transformation in the global AI landscape. The technology is moving quickly from being a commercial product to becoming a strategic asset, a security concern, and a diplomatic bargaining chip.

In the U.S., AI leaders increasingly argue that frontier development is too important to leave to market forces alone. In China, authorities see the same technology through the lens of sovereignty, stability, and industrial competitiveness. Those two visions overlap only partially.

That is why calls for a slowdown now sound different from they did even a year ago. They are no longer just about model safety or ethics. They are also about whether one country can ask another to slow down in the name of humanity while still preserving an advantage of its own.

That contradiction is what made Amodei’s essay so contentious. It was an attempt to move the debate from rivalry toward cooperation, but many in China heard something else: a request to accept U.S.-defined limits under unequal conditions.

What happens next?

The immediate next phase is likely to be diplomatic rather than technical. If the reported preparatory talks between U.S. and Chinese officials continue, the two sides may try to carve out a small set of safety issues that can be discussed without reopening the entire strategic rivalry.

Even if that happens, expectations should remain modest. The history of U.S.-China technology negotiations suggests that progress tends to come in fragments, not sweeping breakthroughs.

Still, the fact that officials and experts on both sides are talking about AI safety at all is notable. A few years ago, frontier AI was not yet a central bilateral issue. Now it is part of the broader conversation about trade, cybersecurity, military competition, and global governance.

For now, Beijing’s response to Silicon Valley’s newest slowdown message is clear: cooperation is possible, but not on terms that look like surrender.

And that may be the central lesson of this moment. AI safety can be a shared concern, but in an era of great-power competition, even the language of caution can be read as a weapon.

Key developments at a glance

• Dario Amodei called for U.S.-China cooperation to pace frontier AI development.
• Chinese officials responded that the proposal sounds confrontational and self-serving.
• Beijing’s main AI worries center on stability, cyber risk, and party control.
• Both countries may still pursue narrow, verifiable safety talks.
• Trust, not technical capacity, remains the biggest barrier to progress.

The coming weeks will show whether AI safety diplomacy can survive the deeper mistrust between Washington and Beijing. For now, China is signaling that it will not endorse a slowdown plan that begins by locking in American advantage.