With so many tech bosses accompanying Donald Trump on his recent trip to China, expectations were high that AI and high tech business deals would be central to the talks. But in the end, there was no breakthrough on selling US firm Nvidia’s high tech AI chips to China.So where does that leave the race for AI dominance and how are China and the US approaching AI differently?Today, China tech expert Selina Xu on who’s ahead right now and what happens next.Featured: Selina Xu, China tech researcher, writer and analyst Subscribe to ABC News Daily on the ABC listen app.Program:More from ABC News DailyChina, United States, AI, Information Technology Industry, Technology, Trade, Trade Agreements, World Politics, Business and Industry RegulationTranscriptSam Hawley: It was meant to be central to discussions, but AI and the race for dominance barely raised a mention during Donald Trump's visit to Beijing. So does that mean China is no longer reliant on America and its AI chips, even though the big US tech giants desperately want to do business? Today, Selina Xu, a China tech expert, on who's winning the AI battle and why it matters. I'm Sam Hawley on Gadigal Land in Sydney. This is ABC News Daily. Selina, last week, when Donald Trump stepped off Air Force One in Beijing, he was greeted, of course, with a lavish ceremonial welcome. But he wasn't alone, was he? He had a few of the big tech bosses with him.Selina Xu: Yeah, I think the most notable few, there was Tim Cook from Apple, Elon Musk, obviously. and then Jensen Huang who joined at the 11th hour.Jensen Huang, Nvidia CEO: Well, I think this is one of the most incredible economies in the world. This is also one of the major nations advancing AI. China's doing great.Sam Hawley: Yeah, Jensen Huang, the head of NVIDIA, he was a last-minute addition to this trip. And of course, he is all about chips, which are the engine of AI. Absolutely.Selina Xu: I think NVIDIA has been one of the most important companies in this US-China AI competition, AI race. And Jensen going, I think a lot of people pre-summit took it as a sign that AI was going to be on the table, that the two countries would talk about export controls and chips. But that didn't really pan out.Sam Hawley: It seems like it became somewhat of the elephant in the room. As you say, the fact that Jensen Huang was there indicated that chips would be discussed, that perhaps business would be done. That didn't happen. But where are we up to in terms of the sale of chips between the US and China?Selina Xu: It's an ongoing situation. It seems currently the fate of NVIDIA chips in China is still slightly unclear. So last December, President Trump, he approved NVIDIA to sell basically its second most powerful chip, the H200 to China. And this is like a major reversal from the Biden era export controls. But since then, despite a lot of back and forth, no H200s have actually been sold in China.Donald Trump, US President: But the H200 is good. China needs it. And so, yeah, it came up. They haven't bought it in so much. No, because they chose not to. They want to try and develop their own. But it did come up, and I think something could happen with that.Selina Xu: So it seems like Jensen was there to try to push this, to make it happen. But it doesn't seem like any concrete progress has actually been made.Sam Hawley: Oh, really interesting. Yeah. So even though Trump, he wound back these Biden restrictions on the sale of these advanced NVIDIA chips, the Chinese have been really slow to purchase them. So why do you actually think that is?Selina Xu: Yeah, it's a great question. The Chinese government has always been, you know, pushing back against export controls, complaining about them and complaining about these broader tech restrictions. But I think a significant thing that happened is how much China's domestic semiconductor industry has progressed, which has offered a lot of confidence to Chinese authorities. So Huawei, I think, is one of the big household names here. They're set to capture the largest share of China's AI chip market this year. This is a huge increase in market share domestically. So if you look at the converse, like NVIDIA, their market share in China's advanced AI chips have essentially dropped from 95% to under 40%. And most recently, I think Jensen was talking about it, to near zero. So this ability of China's domestic companies to fill NVIDIA's vacuum in the time when they've been away, I think this has outstripped most people's expectations. This has kind of changed the calculus in Beijing's head. So now they are trying to push Chinese companies to rely more on homegrown technology instead of just trying to allow them to be dependent on American chip makers like NVIDIA.Sam Hawley: Really interesting that the Chinese at one point did need to rely on US chips, chips made in the US, but now it seems it really does want to go it alone, that it thinks it can perhaps even be superior in this field.Selina Xu: Yeah, I think that is definitely the long-term ambition. And a few years ago, it's almost unfathomable. A lot of people thought Chinese AI industry would not make it. It seems to be destined at that point to be at least like three to five years behind out of the US, which is a gazillion years in the AI industry. And so much has changed.Sam Hawley: All right, so there was no deal done on the chip front. But now why don't we consider perhaps something that's even more important at this point, which is security, because this race for dominance in this sector brings with it risks. Now, some of these new AI models out of the United States, they're increasingly powerful, aren't they? Just tell me about that.Selina Xu: Yeah, incredibly. I think the one that has caught most people's attention is Anthropic's newest model, Mythos. So Mythos is essentially an upgrade from a lot of their previous, very strong coding models with strong coding capabilities and agentic capabilities. But Mythos, what shocked people was how much it could exploit existing operating systems, web browsers. So essentially think of an AI model that could not just do, you know, the usual work that you imagine it to do, but could also do very prolonged, sustained, multi-step cyber attack and cyber defense. So because of that, Anthropic decided to limit the release and only allowed a select group of companies to play with Mythos, to use it to patch up their defenses and not publicly release it until they think enough companies who run, you know, some of the key infrastructure across the economy are able to patch up their defenses.Sam Hawley: OK, so Mythos, it's made by the US company Anthropic, but it's so powerful that even the company is concerned that it could end up in the wrong hands.Selina Xu: Absolutely. Like you can imagine if Mythos level capabilities become widely available, it could mean any person, a lone wolf in a garage or in a cave. It could be a terrorist organization or even a very small country. And you would have the ability to hack major infrastructure systems in the US and in China. So I think that is the fear here.Sam Hawley: All right. Well, the US Treasury Secretary Scott Besant, he's also warned that AI could launch attacks on American banks and perhaps even the nation's power grid. But worse than that, I suppose, is that some scholars have been looking at risks that could see AI design pathogens or actually lead to a nuclear war, an accidental war.Selina Xu: Yeah, these long term existential catastrophic risks are very much of a key part of the AI conversation, the safety conversation, especially in this Trump admin. What's been interesting is Trump and his key AI players came to power and were really pushing for a more deregulatory agenda of letting the AI industry grow. They thought AGI was a bit more far off and they didn't want to talk about safety. Right. There was a very key speech by Vice President J.D. Vance at the Paris AI Summit when he said, we don't want to stifle innovation.JD Vance, US Vice President: We want to embark on the AI revolution before us with the spirit of openness and collaboration. But to create that kind of trust, we need international regulatory regimes that fosters the creation of AI technology rather than strangles it.Selina Xu: And then with Mythos, I think that really was a key turning point. Because of that, the admin, I think, in just the past few weeks have been trying now to figure out how do you make AI safe from, you know, maybe voluntary pre-deployment testing in the same way that, you know, if you before you release like a medical drug, you have to test it by the FDA. So I think that's one thing they're considering. And the other interesting thing, which ties back to the Beijing summit, is Secretary Scott Bessant. He spoke on CNBC and he essentially talked about the need to have a dialogue with China.Scott Bessent, US Treasury Secretary: The two AI superpowers are going to start talking. We're going to set up a protocol in terms of, you know, how do we go forward with best practices for AI to make sure non-state actors don't get a hold of these models?Selina Xu: So I think this is very much now an admin that's coming around to the idea that, OK, we have to consider AI safety and regulation even as we try to encourage AI innovation.Sam Hawley: Yeah, right. So it sounds like the US, because of the advancements in AI and Mythos, is being pushed down the road of regulation because it's becoming too risky not to have regulation. But Selina, tell me about China. We know that China's put massive investment into physical AI, that is robots, which are pretty impressive, let's face it. But what sort of AGI or artificial super intelligence models does it have at this point? Do we know?Selina Xu: Yeah, I would say both countries now still seem a bit far off from reaching this AGI moment. So there isn't a very clear definition for what exactly people mean when they talk about AGI. I think broadly, sometimes people mean the moment of recursive self-improvement, which means essentially the AI can train itself to become smarter and beget ever smarter AIs. Right. So that brings a takeoff moment where you get exponential increases in intelligence. That still hasn't happened. But China, I think what's really interesting is it doesn't seem as AGI-pilled, especially its policymakers. So I think they're very focused on a different set of risks, I would say, than Silicon Valley. I think they're caring a lot now about societal and systemic risks. For instance, AI's impact on labour. They care a lot about child safety and, you know, AI companions. And then also, broadly speaking about, you know, how AI might pose a threat to regime stability. Right. So I think these are kind of the risks that weigh more heavily on their minds. When they think of AI, they don't think of the Terminator or the Matrix. They're thinking of like technology that can help our country get stronger.Sam Hawley: All right. OK, so Selina, this race for supremacy in AI, it does, of course, bring risks. But researchers in both nations, they're very wary, right? Because the concern is you can't win a race when you've got too much regulation. Is regulating this industry effectively a far-fetched dream at this point?Selina Xu: Great question. I think it's at this point increasingly essential. I think the biggest problem with regulation is that a lot of the policymakers who are doing the regulation don't understand the technology. At this point, the hands-off approach is showing that there are cracks. And it's really, really, really essential for the government to figure out the right extent of control. And how do you regulate the various near-term risks and also long-term risks? And trying to figure out, I think, that balance is the key challenge for both the US and China.Sam Hawley: All right. So, Selina, AI was apparently the elephant in the room during Trump's visit to Beijing. The race for dominance, of course, is fascinating to watch it. But what do you think? Who's going to win in the end? Do you think the United States is still well out in front?Selina Xu: The US, I think right now, is about six to eight months ahead if you're just looking at frontier AI capabilities. But if you're looking at it and zooming out in the broader AI competition, the US has the biggest advantage in advanced AI chips in capital. And all of it is being marshalled towards the goal of scaling towards AGI and superintelligence. China, despite being six to eight months behind, is focused on a very different approach, trying to upgrade and embed AI into its various sectors, what they call AI+. And this is being aided by the fact that, one, most of their models are open source, which helps with accelerating adoption, which essentially means that anybody can download the model for free and then modify it to their own means. And then, number two, the fact that most of their models are very compute efficient because they don't have enough chips and therefore they're very cheap. So China's approach might not get the country to AGI and superintelligence as quickly as the US, but it might mean that it's ultimately better for diffusion, for people to use AI across a broad range of sectors. So I think the competition is unfolding along many different lanes and they are running very different races.Sam Hawley: OK, so they're on very different paths. So I guess it depends what you think is the best path to be on at this point.Selina Xu: Exactly. I think you kind of need a combination of both. But the problem, I think, in the US is that people are getting increasingly anti-AI for a myriad of reasons, right? From the fact that they think data centres are spiking electricity and water prices to broader fears about job loss and child safety. So I think people are increasingly against AI and very anxious about what's going to happen. So I think there needs to be a huge narrative change and also some kind of push to make AI feel more visibly beneficial for people's lives instead of just something that's very overhyped.Sam Hawley: Selina Xu is a China tech researcher, writer and analyst based in New York and a former Bloomberg News reporter. This episode was produced by Sydney Pead. Audio production by Sam Dunn. Our supervising producer is David Coady. I am Sam Hawley. Thanks for listening.