Chinese AI companies including DeepSeek and Moonshot are set to brief the UN Security Council on AI safety this week, marking a rare moment of direct engagement on a technology where the two superpowers are locked in fierce competition—and where their competing approaches to regulation could determine whether either nation actually builds safer systems.
The invitation, reported on September 22, comes as Treasury Secretary Scott Bessent proposed a standing China-U.S. AI dialogue in New York just two days ago. Yet the underlying dynamic reveals a troubling reality: the Trump administration's primary answer to calls for slowing AI development isn't to slow it down. It's to make sure China slows down instead.
The Safety-Speed Dilemma
Ten days ago, Anthropic CEO Dario Amodei published an essay titled "We Must Pace the Frontier," arguing that leading labs should deliberately slow how fast their models gain new capabilities. Sam Altman said OpenAI would match Anthropic's commitment to host independent evaluators. Even Elon Musk weighed in: "Dario is right."
But President Donald Trump called fears of an AI catastrophe a hoax. He framed the debate differently: amid the backlash against AI and data centers, "the only one that is happy about it is China." Treasury Secretary Bessent went further, warning there is "no day after tomorrow if China wins" on AI and that "nothing [else] would matter."
This framing has real consequences. A Politico poll found 63 percent of respondents saw a real chance that AI could wipe out humanity. A CBS News/YouGov survey found most Americans want AI development slowed, though not stopped. Yet when asked whether falling behind other countries is acceptable in exchange for safer AI, Americans split 43 percent to 37 percent—and Trump's 2024 voters preferred staying in the AI race, 50 percent to 38 percent, over prioritizing safety issues.
The administration's strategy is to maintain speed at home while constraining China's access to the technology. The Trump administration has cleared roughly 10 Chinese companies to buy Nvidia's H200 chips, though Chinese customs has kept nearly all the chips out. The administration also sees tighter controls on chips, compute, and access to American models as a possible way to slow Chinese AI instead.
Where Institutional Safeguards Fall Short
Amodei's essay outlined a more nuanced approach. He suggested "an attempt to coordinate with authoritarian governments, to the extent this is possible, while taking seriously the challenges of verifying compliance," with Beijing at the top of the list. He also urged keeping "powerful AI chips or semiconductor manufacturing equipment" out of Chinese hands, closing "chip smuggling operations" and remote access to data centers, cracking down on unauthorized distillation, and hardening labs against theft of model weights.
Yet Bessent testified before the House Financial Services Committee seven days ago and rejected the liability exemption the labs are seeking, arguing that safety is best guaranteed when developers are "liable for what they build and generate." The independent evaluators that Anthropic and OpenAI have agreed to host don't require the United States to fall behind anyone. They're a form of democratic oversight—a way to make companies answer for their products.
That's the missing piece in the administration's approach. Treasury Secretary Bessent said in July that Washington could sanction firms found stealing from American models. On September 15, the NSA, FBI, and CISA jointly accused six Chinese developers, including DeepSeek and Alibaba, of industrial-scale distillation of American frontier models. But the administration's response has been to double down on export controls rather than strengthen the domestic accountability mechanisms that might actually make AI safer for everyone.
The Diplomatic Challenge
Xi Jinping's arrival in Washington this week marks the first state visit by a Chinese leader in more than a decade. Trump said he wants to stay ahead of China because "whoever wins AI wins." The summit is expected to shape policy on export controls and diplomatic signaling.
Two of China's most senior AI governance figures—Xue Lan, who chairs China's national expert committee on AI governance, and Zeng Yi, dean of the Beijing Institute of AI Safety and Governance—joined a Capitol Hill discussion convened by Senator Bernie Sanders this spring. Xue called for "safe zones" where American and Chinese scientists can build safety standards and protocols together. Zeng warned last week that AI's catastrophic risks, including "the hidden danger of human extinction," are mounting daily.
But there's a credibility problem. The AI consultations Trump and Xi discussed in Beijing in May were never formalized. China's Foreign Ministry answered Amodei's essay by warning that "fearmongering, confrontation and vicious competition" would disrupt global AI governance. The state-run Global Times called the slowdown push a "Cold War playbook" for the sector. China's Commerce Ministry threatened "resolute countermeasures" if Washington uses the distillation accusations to suppress Chinese firms, and China could resurrect rare earth exports controls that were suspended under the Busan truce in November.
China-U.S. crisis management mechanisms already have a poor record. The February 2023 Chinese high-altitude balloon incident led Beijing to later explain that Washington had not created "the proper atmosphere" for dialogue.
Amodei argued that restrictions "increase the leverage held by democracies and make an agreement more likely in the future," and that sharing information about misaligned models can have value even without formal agreements. The Trump-Xi summit could produce a first substantive commitment that neither side will use AI to produce biological weapons. The two sides should establish a notification mechanism as a standalone safety channel, insulated from trade and technology-control disputes, and pair it with a technical track involving the Commerce Department's Center for AI Standards and Innovation and experts like Xue and Zeng.
Why This Matters:
The administration's strategy—forcing China to slow down instead of slowing down domestically—may play well at home. But it won't make American AI safer and risks scuttling a China-U.S. channel on AI risk before it begins. When 63 percent of Americans believe AI could pose an existential threat, and when the two countries with the most advanced AI systems can't even formalize basic safety consultations, the public interest isn't being served by export controls alone. Domestic accountability—through liability frameworks, independent oversight, and democratic scrutiny—matters as much as international competition. The Trump-Xi summit will reveal whether either side is willing to prioritize safety over speed.