Hey everybody, my name is Nick. Welcome to the NDS show where we discuss critical topics in national security, intelligence, and geopolitics. Joining me in a moment is Dmitri Alperovitch, fresh off his testimony in front of the Senate Select Committee on China. We have an amazing show for you today, including Russian hacking of your home internet. Recently, Russia was busted hacking Chinese-based TP Link Routers, which would give them access to any information traversing through your home. All of this, plus what you can actually do to protect yourself in the future. But first, the AI arms race with China heats up. Are we moving closer and closer into World War II? And are we giving away the nuclear weapon of our time in artificial intelligence? Before we hop into it, be sure to subscribe on YouTube and follow along on X and TikTok. Let's go.
The question is, should we be shipping our AI infrastructure to China? Should we allow them to use Nvidia chips and all our hot tech?
Yeah, Nick, to me, this is a complete no-brainer, right? I mean, just ask yourself the question, the last time we were in a great race with a superpower, that was the moon race, right? in the 1960s. Imagine making the argument, let's sell rockets to the Soviet Union because that will help us get to the moon faster. That is literally the argument today. It's actually even worse because the stakes are so much higher than getting to the moon because artificial intelligence is going to determine who's really going to win the race for the 21st century. Not just the AI race, but you know, it's about economics. It's about military power. Literally every model that China is going to build is going to get integrated into their military systems, into their intelligence systems that's going to be done in a way that will hurt Americans. In fact, you know, even if you don't believe that we'll ever get into war with China, and hopefully we won't. You know that those weapons that they make are being sold to our adversaries. They've sold weapons to Iran, for example, right? So imagine those weapons being so much better that they're going to help kill more Americans. Like literally, those are the stakes. And the only arguments I hear in favor of selling chips to China are coming mostly from Nvidia that just wants to make money. But the arguments that they're making is we're going to get China addicted to our chips, which is so ridiculous because chips are not cocaine. They're not addictive. In fact, you look at all of the frontier companies in the US, they're using a variety of different chips. They started with Nvidia, most of them, and then they moved off to Trainiums from Amazon, TPUs from Google, every single hyperscaler, every single frontier model is now working to design their own chips effectively. And they're doing it super quick. So there is actually no addiction. And the other argument that you hear is well if they buy our chips, they will not build their own, which is just so ridiculous because of course they're going to build their own. Anytime they hear that we're going to get them addicted to our chips, they're saying wait a second, we need to make sure we invest in our own infrastructure. And that has been a long-standing goal for Xi Jinping who has poured billions, hundreds of billions of dollars into the sector. He does not want a dependence on US designed chips. They're going to keep investing in their own production as you would expect them to do. And you're going to end up with more chips in China, better models that are helping their military, helping their economy, and helping them build overall better models. Like why in the world would we do that?
I think artificial intelligence is the crux of the situation with China right there. Like this topic is going to decide the future on our relationship with China for a few reasons. It is the perfect cross-section if you will between economic and national security concerns. It's the pinnacle of those two things.
What are the national security concerns? And then do we allow our economic concerns to take precedent? Like that seems...
We don't actually have economic concerns because it would be one thing to say Nvidia is really struggling. You know, they're about to get delisted. They need the market to survive. That is not the case. It's one of the world's richest companies in terms of market cap and frankly in terms of revenue. And by the way every single chip that comes out of the TSMC production line for Nvidia has a customer in the United States. Literally you have hyperscalers, frontier companies saying we can't get enough chips. And instead of saying, 'Well, we're going to supply all the chips you need,' Nvidia is saying, 'No, no, we're going to take some of that production and we're going to ship it to China.' Literally, this is a zero sum game where every chip that goes to China is a chip that does not go to a US hyperscaler. So, this is not even an economic concern of like, oh, we're going to make more money by selling chips to China. It is literally diverting demand to China that exists here in the United States.
And how many of these chips are going to China? Like what's the scale of this?
Well, the United States has approved licenses for the H200 chips. And apparently they've approved some of them to export them to China. And we're talking potentially as many as a million chips that could go to China, which is, you know, still a small fraction of what gets sold in the United States. Sure. But that's still a lot of chips that can help China along with their indigenous chips, the Huawei Ascend chips to build really good models. And if you look at why the Chinese models are so close to us, about 12 months or so behind, there are really two reasons. I've talked to many leading researchers that have confirmed this. One is they've been able to procure a lot of chips, some legally, some illegally. And two, even though they have less compute than we do, they're distilling our models, right? Distillation is basically when you create tons of bot accounts on Gemini, on Claude, and ChatGPT and you send lots and lots of prompts, literally millions of prompts into these systems and collect reasoning outputs and then use it in the post-training models, right? So that allows you to basically steal the reasoning from US models, use less compute. You still need compute, but you need less of it to get better, right? And that's why they're sort of perpetually behind, but fairly close behind. So, we need to address both. We need to sell them less chips to make this post-training less effective, and we need to work against distillation to prevent theft. And by the way, that's another reason not to sell them chips. They're stealing literally IP. Why would you sell them chips while they're doing that?
Steal IP and then here's some more of our best technology in the world. It's incredible. Like it just makes absolutely no sense.
It's interesting to think about this topic in regard to who's in power right now in the United States because it's been pretty clear that we have a commander-in-chief that's willing to act on these sorts of things. And if I'm China, I'm saying, 'Yeah, this is a national security concern for us. The Americans can't have the best AI. We have to have it.' And they're less scrupulous, right? They'll break the law. They don't care necessarily about law to get to the best AI or whatever. So, they'll rip it off. They'll steal it just like you're saying. What's a proper response from the United States? We just say you don't get any chips. And then do they say, 'Well, we need those chips.' You know, what's the response here if this really is a zero sum game?
Yeah. Well, look, Nick, I just testified on this very thing as you said in the House in front of the China Committee, and I basically said that it boils down to two things. One, let's not arm the adversary, right? Let's not sell them chips. Let's make sure they don't get access to our chips even when they're deployed in cloud outside of China because, you know, even if not a single chip is going to China, but they're using chips in clouds in Malaysia and elsewhere, that still is a huge problem, right? So, let's close all these loopholes. And two, let's not create another Strait of Hormuz, right? Let's not create more choke points than we would have on an adversary like we currently have on oil coming out of the Gulf that Iran is trying to block. Let's make sure that all of our choke points with regards to China, and they relate to foundational chips that they're producing, they really relate to obviously critical minerals and particularly rare earths and pharmaceutical precursors and a few other things. Let's make sure that we reduce those vulnerabilities, that we're not relying on China in case, god forbid, there's a conflict over Taiwan or South China Sea or whatever. And this is going to be important to make sure that we have the freedom of action to do what's necessary to protect our national security interests without thinking, oh, how is China going to retaliate against us economically?
Speaking of national security interest, I saw recently there's a company named Miservision. Have you heard about this? So a company named Miservision was acquiring... it was actually the select committee on China that put out this tweet. Miservision was acquiring satellite imagery from Airbus.
And sharing that with the Iranians. Now, that's a direct national security threat. That's black and white, right? You're sharing this with a current combatant that we're fighting in a war. How does that change the dynamic of our policy towards China? And should it...
That's been going for a long time. China has been supplying precursors for chemicals for producing missile fuel which they've been using to fire all over the Gulf and kill Americans in this latest conflict. Right. They've been supplying parts for different electronics that go into drone systems. They've been supplying them with air defense, including apparently there was an FT story, I believe last week, that they're going to supply them with thousands of MANPADS, shoulder-fired air defense weapons. One of them was used to shoot down our F-15 over Iran a few weeks ago and China is going to resupply them during the war. So this is a country that at a minimum is supplying our enemies, by the way helping Russia in its war against Ukraine as well, and at a maximum is building a military that is going to challenge the United States in East Asia. Right? So why you would do anything to help their military-industrial complex, their economy, because artificial intelligence of course is going to be critical to economic growth in this century is just beyond me.
In the intelligence community and probably national security community, you always often hear China referred to as the global competitor. They're the competitor. I mean, the way that you're breaking this down, it doesn't sound like a competitor to me. It sounds like they are the enemy.
They are the adversary, right? I hate that term competitor. I hate that term competitor because it assumes that this is sort of a boxing match, right? and we have rules which by the way China does not abide by any rules and two like you lose, you know, there'll be another match right? doesn't really matter. Well no, the stakes here are quite existential. The stakes are who is going to be, whose century is this going to be? Is this going to be another American century like the last one was or is this going to be China century where China dictates the rules of the global order, economic order, security order in the century and we're going to be living in a Chinese world. I don't want to live in a Chinese world. I want to make sure that the United States remains the world's preeminent superpower and anything that we can do to weaken them we should be doing. That's my view. I believe we're in a new cold war. One that is actually remarkably similar to the last one. Right? You look at all the elements of the last cold war are essentially present here. You have the global competition for supremacy playing out in every single corner of the world. Not just in East Asia but in Africa and Latin America and Europe where China is trying to extend its influence even here in the Caribbean. You have an arms race, both countries arming themselves specifically for the conflict with each other, right? When the president talks about he wants a $1.5 trillion budget, China is literally the reason to do so, right? It's not Iran, it's not Russia, it's not any of these other countries that we may have problems with. It's China. And China, of course, is undergoing the biggest military modernization in human history. And by the way, the arms race is nuclear as well because China is building up its nuclear stockpiles, nuclear weapons stockpiles. We're modernizing our arsenal because of China. We have a tech war, right, playing out in artificial intelligence and chips and many other things. We have a spy war where the FBI says the level of espionage against us is the highest it's ever been, including during the first cold war. Most of it from China, a lot of it enabled by cyber. We have an ideological fight, right? A communist authoritarian country versus democracy, a capitalist democracy. So you have like literally all of those things that we had. And by the way, we even have a space race, right? We're trying to get to the moon, hopefully landing a man on the moon in the next couple years before China does, right? That's like the urgency is like we got to get there before China does. Like if that's not familiar to you from the 1960s, I don't know what is. Literally like it's basically the same cold war as we had with the Soviets except that China is actually much weaker than the Soviets ever were. They're much more constrained geographically. They're literally surrounded by countries that they have problems with. Whether it's Japan, whether it's the Philippines, whether it's India on the other side, even Vietnam, they fought a war with them and literally have border conflicts with all these nations. They settled one with Russia about 25 years ago or so. And they don't have the Warsaw Pact. They literally have no allies, right? They have friendly countries like Venezuela, like Iran that they didn't lift a finger for when we, you know, kidnapped basically the president of Venezuela and attacked Iran. So the only country that they have a mutual defense treaty with is North Korea. Not exactly world superpower. So, you know, they're very weak. We have all the strengths here in AI. We're clearly winning. But yet, as my friend Niall Ferguson, preeminent historian, says, there's nowhere in the history books that it's written that America always has to win cold wars. And we're perfectly capable of screwing it up. And one of the ways in which we screw things up is by deciding, well, let's give away our advantage to make a buck. Let's sell our technology to China. By the way, we did this with the Soviets. People forget this. I write about this in my book World on the Brink that during the 1920s and 1930s we sold all sorts of technologies to the Soviet Union. We built their steel plants, we built their automotive plants that were later converted to tank and airplane productions. And then we woke up one day and we're like, 'Oh my god, maybe they're actually an enemy of ours.' And we helped build literally their entire industrial sector, right? And that's what we've done with China. You know, they would not have the economic power they do today if we didn't open up China in 1979 with Jimmy Carter deciding to revoke recognition from Taiwan, recognize China, open up trade. Our capital, our technologies went into China, helped build massive industries. We wonder why China is so good at electronics these days, microelectronics, because Apple went in in the early 2000s and decided to outsource all their production to China and build a massive industry there and provide them with all the skills that they now have evolved on and have the best arguably manufacturing, microelectronics manufacturing production in the world. Like we've enabled it all and now we're waking up and saying, 'Well, wait a second, maybe that wasn't so smart.'
We built the monster. Now we got to fight the monster in some regards like again and again.
Yeah, it's true. I mean around the world too, just not just China.
You spoke a little bit about Russia in there about you know being allied with Russia. I mean China and Russia are working hand in glove in certain respects. Does that alliance need, should it be a main goal of ours in the United States to try to break that apart and how do we properly approach that to gain leverage over China and Russia at the same time?
Yeah, I don't think you can. Look, this is not an alliance, right? They don't have a mutual defense treaty. This is at best a friendship of convenience where China is not supplying direct weapons to Russia. They're supplying a lot of dual use capabilities that are very essential for continuing its war against Ukraine, but it's not providing direct weapons. And everything that China is doing with Russia is for its own benefit. It is extracting economic price from Russia for these technologies. It is getting more military technologies that it needs for the invasion of Taiwan from Russia, like airborne training and technologies that it doesn't currently have that Russia is supplying to them. And look, you're not going to break up this friendship because both countries have an overarching goal, which is to take down the United States, right? To make sure that this is a multipolar world, that US power is constrained. That is the top priority for Russia. That is the top priority for China. That's what's bringing them together, right? Obviously, that is not a priority for us. And that's why you're not going to be able to do this, you know, fanciful reverse Kissinger of like become friends with Russia in alliance with China. Russia has no interest in that.
Doesn't that kind of seem like what's going on? Like it seems that we've started to support, not support, but we've not been as friendly with Ukraine recently. At least in the media you could say, I don't know behind closed doors that might be different. And it seems like at least our leaders have shifted more towards okay let's talk with Russia, let's work with Russia, let's bring Russia to the table. But yet nothing's happened there. What are we doing wrong?
Well, look, you know, the president clearly wants to end the war in Ukraine. And you know, that's an admirable goal for sure. Like this war has dragged on. It's now in its fifth year. You have hundreds of thousands of people that have been killed, many more maimed. You know, many civilians died, lots of infrastructure destroyed in Ukraine. So it has been, you know, a huge tragedy. You know, the problem is like, you know, we've tried for a year to offer all sorts of proposals to Putin. He's not interested. I don't think he wants to end this war while Ukraine survives as an independent sovereign nation free of Russia's control. His goal is to subjugate Ukraine at least politically, if not completely from a sovereignty perspective. And he's not interested in anything less at the moment. He still thinks he can accomplish this on the battlefield. So he doesn't need a deal. You know, we will see what happens as the Russian economy starts to struggle more, as the recruitment numbers are no longer able to compensate for the losses on the battlefield. Since about December, their losses have been exceeding new recruits coming into the fight. So this is not sustainable for a long period of time but while he still thinks that he can win this he's going to keep going.
So something kind of amazing happened recently in Ukraine at least, at least this is how they're promoting it, and that's an entire enemy position was taken through autonomous weaponry through drones and robots. How does that shift the dynamic there? Or does it shift the dynamic there? Does it become a manufacturing war? And does Ukraine have the advantage in that regard?
Well, it has been a manufacturing war as well as a personnel war. You still need people to launch these drones, to, you know, taking positions is one thing. Holding them is a different story entirely. So, you still need people for that as well. So, you know, obviously Ukraine is trying to hype up its capabilities and they are impressive, but this is not a war where you don't need people, right? And that remains a critical shortage that Ukraine has and that's why it's steadily losing ground, right? So, you know, for all of the pain and attrition it is inflicting on the Russian forces, it's losing ground. Not huge amounts of ground and the rate of advance of Russia has slowed down dramatically but it's still, you know, every single month there's more and more territory that's being lost right and a lot of it is because they don't have the manpower to defend across the entire front line. But drones, whether it's aerial drones, ground vehicles, maritime vehicles, absolutely essential. Russia of course is now in this fight as well and they've done actually some of the innovations like the fiber optic drones where you have a fiber optic cable dragging on for miles behind a drone that allows you to fly and not expose any radio signal, right, that would be jammed or detectable. You also have an ability to do different types of tactics with drones because you consume a lot less power. So, a drone that's supplied with fiber optics cable can survive for days. And they're basically using them as sort of movable deployable mines where you fly to a location, you land, and you wait hours or days for a truck to go by, a resupply vehicle, and then you hit it right as it's going by you. So, a lot of innovation on both sides are taking place. No question that drones have become essential in warfare. We just saw this in a conflict with Iran where so many Shahed drones have been sent against Gulf states and against our targets as well. So, you know, we are trying to build our own obviously and it's been interesting that the Lucas drone that we used against Iran was actually a reverse engineer basically that we sent back at them. So, we are, you know, it's a good indication we're not so proud that we have to engineer everything ourselves. We're building and designing our own drones, but we're also saying, 'Hey, that one is pretty good in terms of low cost munition. Let's just use what they've got.' And that was a drone that we captured on the Ukraine battlefield. Ukrainians provided to us and reverse engineered and used against Iran.
I think what's happening right now is this. We're in the middle of this shift towards fully autonomous warfare. And recently in the proposed fiscal year 2027 NDAA, there was something kind of hidden in there that you got to kind of look for it, but they're increasing a unit called the Defense Autonomous Warfare Group from 225 million to $54 billion. And this group, all they do is, well it really it sounds like the establishment of an autonomous warfare command is where this is going. But this group is focused on the long range drones, the big drones, not the small FPVs. And that shift is happening under our feet right now. If that gets approved, 54 billion into autonomous weaponry, how does that change the game globally? I mean, if imagine for a second, we have this ongoing issue with Iran if we didn't have to send any humans in there. We just sent in autonomous weapons and said, 'No, you are going to open the Strait of Hormuz because I have the T2000 staring you down right now.'
Well, look, no question that autonomous weapons are going to become more and more prevalent on the battlefield. Again, I would caution you against thinking that that means that humans no longer have to fight wars. That will still be the case for the foreseeable future. Wars will involve people. People will die in wars. This is a capability that augments manned aircraft, manned vessels. It does not replace them, right? So, we're still going to need aircraft carriers. We're still going to need F-35s. We're still going to need lots and lots of things, infantry. That is manpower, right? And hopefully these capabilities are going to make them more lethal, more capable. But it's not going to replace people anytime soon.
Do you think it makes war more likely, more trigger happy? If you don't have to send in a human somewhere, if you say, 'I'm just going to send in this drone swarm to take out your leadership.' Does that make war more likely?
Well, anytime you can launch war at relatively little expense in terms of blood, you know, you open up options, right? I mean, look at this conflict with Iran where we've inflicted thousands of casualties on the Iranians and lost tragically but lost 13 Americans, right? An incredible asymmetry here and basically owned their skies and were able to bomb at will. And you know one of the incidents was a tanker refueling accident, not even a combat set of fatalities. Tragic nevertheless. So, you know, anytime you're able to sort of execute a war on the cheap like that, you know look at the Venezuela operation right we didn't lose a single American in that fight although, you know, we sent in people and there was a very intense firefight. Autonomous weapons obviously can help reduce casualties potentially but look, you know, our adversaries are procuring them as well. So you look at what Iran was able to do in terms of launching a steady number of missiles and drones at allies around the Gulf and at Israel basically unabated despite our overwhelming advantage in Iranian airspace. You know, it tells you that the adversary always has a move to play and we need to be careful.
I want to shift towards maybe more of your background in the cybersecurity space. Formerly with CrowdStrike, is that right?
So recently Anthropic soft launched a new model called Mythos that they're enabling all the big companies right now to use it to patch up their old software and things like that. Because if they release it to the public, they were fearful that it's so good at coding that it can break into anywhere. It can break into anything.
In the cybersecurity space, how does this change things? This type of technology that can work autonomously and break into anything. I mean, it sounds like it can break into anything.
Well, what it can do is find vulnerabilities in most software packages. And by the way, that model may never get released. We'll see. But they are working with...