Everyone is talking about the spectre of AI catastrophe. Can the U.S. and China agree on guardrails to blunt the biggest risks when their leaders meet next week?
Transcript
MARY LOUISE KELLY, HOST:
If you are like me, you have spent this past week hearing a number of AI-driven doomsday scenarios. Warnings about the extreme risk of superintelligent AI have been pouring out of the tech world. President Trump has largely swatted away any and all words of caution about artificial intelligence, calling them a hoax. But there is one AI scenario that concerns him – falling behind China. I asked NPR White House correspondent Emily Feng about the possibility on our weekly national security podcast Sources & Methods.
(SOUNDBITE OF EMANUEL KALLINS AND STEPHEN TELLER’S “PLUNGED INTO CRISIS”)
EMILY FENG, BYLINE: President Trump has directly cited one of the reasons why he supports AI expansionism as, basically, we don’t build this stuff first, no matter how dangerous it could be. China is going to do it first. And there’s been a very robust debate about, well, can China actually build that first? The U.S. still has a pretty substantial lead when it comes to hardware and software abilities when it comes to AI. So can it build these frontier models faster than the U.S.? And actually, does it even really want to? You know, it’s clear from Chinese state planning documents they want to develop AI capabilities, and they want to apply that as fast as possible. But arguably, it’s not in their interest to build a model that then could turn around and destroy all of humanity.
And then there’s a second parallel discussion that I’ve been following about there’s an obvious interest in the two leading countries that host the most powerful AI companies to come together and talk about broad safety guidelines. But is there any real political incentive for them to do so? And you see President Trump and other people, his advisers around him, arguing, this is a risk that we cannot take. If we voluntarily constrain ourselves, China might promise to do something but then flake out, and they’ll use that time to then surpass us.
KELLY: A couple of pieces to follow up on in what you just said. One, you were talking about the people who advise Trump. Who is he listening to on AI that is informing his thinking about what the risks are and are not?
FENG: Trump has surrounded himself with a lot of technology CEOs. There’s been a lot of reporting about his fond relationship with the CEO of Nvidia, which makes a lot of the graphic…
KELLY: Chips, yeah.
FENG: …Processing units, the chips that go behind data centers that power AI. Trump has flown with Jensen Huang, the CEO of Nvidia, to Beijing. And so Huang, at Nvidia, has been a big proponent of trying to accelerate the development of AI in the U.S. as quickly as possible.
You also have Trump’s former AI czar, David Sacks, who is a big-time investor in Silicon Valley, who’s written a lot on social media dismissing all of these concerns about AI as doomerism. And so those are the kinds of voices that are influencing Trump right now.
And the other, I would argue, is the U.S. economy. There’s going to be about $1 trillion of investment in data centers this fiscal year alone. So AI is one of the things that’s propping up the U.S. economy, which we know the president really cares about, despite things like the war in Iran and inflation.
KELLY: I think you’ve just started to answer what I was going to ask you next, which is, how do we square legit concerns about not wanting to get left behind by China or, you know, let them race ahead of us on this. And yet, this White House has greenlit the sale of advanced chips to China, which they need to develop AI.
FENG: Yes. Trump is very pro-business. He wants more sale of American chips abroad. So he’s reversed decisions that would limit the sale of advanced semiconductor chips to China. And if we want to beat China, but we’re also selling the same hardware that allows them to beat us, that’s kind of a big tension in your AI policy.
The other assumption that I want to dig into, though, is that regulation would necessarily slow down the development of AI. And I think weirdly, China is a good example of this. So China put very strict domestic controls on AI development and public usage, mostly for politically self-serving purposes. A lot of the strictest rules are about what kind of content can be made available to the public. And yet, we’ve seen Chinese companies make really, really fast developments to the point where they’re almost catching up to American AI companies. So it’s not necessarily the case when we look at China that having regulations in place limits the development of AI.
KELLY: Do we know – and I’m asking you to put your former China correspondent cap back on – do we know how this conversation is playing out in China? Are they also having a raging debate about AI disaster and how imminent it may or may not be?
FENG: They have been talking about this for a long time. What has really captured people’s attention has been Dario Amodei’s paper arguing for slowing down AI. And specifically…
KELLY: That’s being read and discussed in China.
FENG: Very much so.
KELLY: OK.
FENG: And specifically pointing out China in that essay and saying China is our pacing challenge, and we need to ask China to also voluntarily slow down their development. And that’s been very controversial because a lot of people in China have said this is disingenuous. This is a hoax to get us to slow down while the U.S. then doubles down and runs even farther ahead of us. And so I think that demonstrates, again, that although there’s a lot of interests politically in China for some kind of global regulation of AI, there’s also a complete lack of trust as to what the U.S. government and U.S. companies’ ulterior motives really are.
KELLY: That was White House correspondent Emily Feng on NPR’s national security podcast Sources & Methods. Our cybersecurity correspondent Jenna McLaughlin also joined us for a wide-ranging discussion about AI. You can hear the whole thing wherever you get your podcasts.
Copyright © 2026 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.
Accuracy and availability of NPR transcripts may vary. Transcript text may be revised to correct errors or match updates to audio. Audio on npr.org may be edited after its original broadcast or publication. The authoritative record of NPR’s programming is the audio record.
