Project Maven, the Dawn of AI Warfare

0:28They told us to expect change. The thing about AI for business, it may not automatically fit the way your business works. At IBM, we've seen this firsthand. But by embedding AI across HR, IT, and procurement processes, we've reduced costs by millions, slash repetitive tasks, and freed thousands of hours for strategic work. Now we're helping companies get smarter by putting AI where it actually pays off, deep in the work that moves the business. Let's create smarter business, IBM. Being a small business owner isn't just a career, it's a calling. Chase for Business knows how much heart and effort go into building something of your own.

1:08Manage all your business finances, from banking to payments to credit cards, all in one place with Chase's digital tools. Plus, access online resources designed to help your business thrive. Learn more at chase.com slash business. Chase for Business. Make more of what's yours. The Chase mobile app is available for select mobile devices. Message and data rates may apply. JPMorgan Chase Bank N.A. Member FDIC. Copyright 2026. JPMorgan Chase and Company. Bloomberg Audio Studios. Podcasts. Radio. News. You're listening to Bloomberg Businessweek with Carol Masser and Tim Stenevek on Bloomberg Radio.

1:50So the U.S. war in Iran is now about one month old. Attacks from all sides continue, as do clashes over peace talks. We've talked about the cost of war, lost lives, destruction in the Middle East region, the impact on global energy markets, billions of dollars spent on military equipment, and the changing face of how war is fought. For the United States, that has included the role increasingly of artificial intelligence. CENTCOM Chief Brad Cooper addressed that in an update on the U.S. war in Iran. This was back on March 11th. Our warfighters are leveraging a variety of advanced AI tools. These systems help us sift through vast amounts of data in seconds so our leaders can cut through the noise and make smarter decisions faster than the enemy can react.

2:33Humans will always make final decisions on what to shoot and what not to shoot and when to shoot, but advanced AI tools can turn processes that used to take hours and sometimes even days into seconds. And then a CENCOM chief Brad Cooper back on March 11th. AI in War, It Is Here, Algorithmic Warfare, AI for Intelligence and Weapons, We might be able to link it all back to a vision, an idea written in a thesis by Drew Cukor back in 1997. He wrote, white dots fused with an understanding of the enemy and validated by other collection assets can contribute to a highly accurate battle space picture.

3:08It's a lot. But this seemed to be his mission and may have led to what some have described as the most important secret military project of the 21st century. She has been covering this secret for a long time. It's all in a new book, Project Maven, A Marine Colonel, His Team, and the Dawn of AI Warfare. The author, of course, is Katrina Manson. She's a Bloomberg News reporter covering tech and national security. She's here in our studio. Congratulations. This has been a long project for you. And we want to get into the book and who Drew is and what he did. But tell us, for those who maybe still don't know, what Maven is, what this system is.

3:44There are two parts of it. what the US is using right now in its operations against Iran is called Maven Smart System. It emerges from this project that starts in 2017. But what it is today is if you imagine something like Google Earth, Google Maps for war, I imagine a digital screen and a map, but on it are more than 150 data feeds coming in, AI helping to sort through those feeds, computer vision helping to identify possible targets. And what has happened particularly in the last year or so is AI tools, LLMs, large language models, helping to begin to offer courses of action, pairing even ideas for pairing a weapon to a target.

4:33And that is what is helping speed up those processes that the Admiral was just talking about. And we've used it already. And by we, I mean, the Department of Defense, the Pentagon, out on the battlefield. Carol mentioned 2017 and the genesis earlier from the 1990s with Drew Cukor. But I'm also wondering about where in recent years, the technology has been used. The first time it was really put through its paces is in this very difficult moment in US military history in the withdrawal of US troops from Kabul in Afghanistan in 2021. And at that point, there was a very difficult situation, people coming into the airport, I'm sure all the viewers remember those scenes.

5:20Computer vision was used to try and count the number of crowds. Now, anyone could see with the naked eye, there was a crisis unfolding. But this computer vision was able to pick up very quickly how many people were there, and send it back to headquarters and help the people on the ground really emphasize how bad the situation was. It was really the first time that multiple people could be on a system without it crashing. So a lot of this is not so much about AI, but bringing data and systems together, networking and connecting people on the ground to people back in the Pentagon. I want to go back to the origin story that you get into.

5:58And so many different things came to mind. I think about, you mentioned like 9-11, right, and the data that was missed. But it was there on government computers, but things not talking. And the role of Drew and what he wanted to do, because bringing data together to maybe make the military smarter, maybe save more lives, maybe be smarter in terms of how it does things. Tell us about Drew and who he was. He's this fascinating Marine Corps colonel by the end of his career. But at the beginning in the book, I joined him as a very young intelligence officer. who is angry really from the start because he doesn't have the right tools.

6:34And he's very enterprising. He is on the internet right at the beginning reading open source newspapers, but he manages to run up this huge bill by connecting his computer and gets in trouble. So he is constantly someone who the bureaucracy is saying, don't overspend, use the existing tools, stay in your box, and someone who is constantly refusing to do that. And he develops this grand vision to bring more intelligence to operators. Rather than just leave it within the intelligence function, he wants to get information onto the front lines where troops are risking their necks. Well, what was it about his specific experience in war that motivated him to keep pushing for this?

7:16There are a few. One is in October 2001, he is part of the Marine Expeditionary Force sent to Afghanistan. And so he is an intelligence officer. And of course, you'll remember that very quickly the Marines are being maimed and killed by improvised explosive devices. Finding out where those were, the patterns that the Taliban were using, even the weather systems that meant that bomb layers were more likely to lay weapons, was something that really eluded the U.S. for years. And in 2011, he's back in the Pentagon, and he tries to get this company that is now very well known, Palantir Technologies, to bring their data analytic platform out into the field.

8:02Now, the U.S. already had platforms like this, but the common complaint was they weren't fit for purpose. So I find, first of all, I feel like I think about Katrina, where we are in AI, and I think how much actually the government 10 years ago, like either working with Palantir or understanding how valuable data AI specifically could be in warfare. And I just think about how that maybe got everything rolling. And like, go back to their reach out for Silicon Valley and how Palantir came into it. Like, was Palantir just, I know it came through a certain individual, but just kind of at the right place at the right time?

8:43It's actually even more happenstance than that. Drew Kukul wants Google. Everyone knows that Google is great from a sort of access to the mapping facility and the CIA. Which the government was already, right, involved in. Yes, and In-Q-Tel, the CIA's investment arm, had invested in the company that becomes Google Earth. So they had a real sense that this was the platform they'd want to use. But when Project Maven started in 2017 and Drew Kukor becomes the chief of this project to bring AI computer vision to drone footage to pull out data that they couldn't currently get, what's been publicly well covered is that Google protesters found out that they were on this project and were not keen to work with the Pentagon.

9:31When Google responded to this crisis, they decided not to renew the contract. They decided that they wouldn't work on weapons. And Drew Cukor has a problem. He needs a cutting-edge technology company to deliver his vision for war. And he rings up Palantir. And he says, hey, have you got a minute? And he flies over and meets with a team and says, this is my vision. And actually, Palantir doesn't want it. They are not an AI company. They are one of the AI skeptics at this time. And they also don't want to make a user interface. They don't see themselves as the sort of fancy, frilly, nice-looking bit.

10:13They want to do the data crunching. They concede because Drew Kukor is a very compelling, convincing person who makes his case. And slowly this partnership develops really between Aki Jain, a senior official at Palantir, and Drew Cukor working late nights to develop this platform to deliver what is now being used in the future of war today. Pretty amazing. I want to talk more about the companies that are involved in Project Maven, but before we do that, I want you to talk a little bit about Drew Cukor personally and just the chasing that you did to get an interview with him, the first interview you did with him, where it was, the realization that you had during the interview.

10:55Because it was so easy to get, right? I mean you spent what a year trying to talk to this guy and he finally agreed to talk to you yeah take us through that I had already written for for Bloomberg a piece about Maven smart system and its usage by the 18th airborne Corps and this piece had come out and there had been reaction by people who knew Maven very well and they'd said well you know you really need to speak to someone else you haven't spoken to yet who is Drew Kukor you can't tell the story of AI warfare without him. And eventually, after a year, he did agree to meet and he was working at JP Morgan.

11:30And we met after work. And he offered me water, he took nothing for himself. And we sat opposite each other. And I very quickly realized that he was interviewing me. And he is a very interesting person. He's, he's compelling, he is by all accounts, extremely hardworking, and for some people extremely difficult to work for. In the book, one person describes him as a psychopath and then adds, in the best way. And so I have spent quite a lot of time trying to work out what that means. He didn't really eat at work. He didn't drink. He just worked and he expected everybody else to do the same. And as he later tells me, we end up speaking every week.

12:17And I began to really look forward to Friday afternoons, which was our time, because I was learning so much about a field that is so complicated, and an issue that is so fraught that has divided people. Well, that's what I want to get into. But I want to talk a little bit more about the project and the people who worked on it, the Mavenites, you know, who worked on it, there was there was one thing we read, the interview process, and one individual mentioning something about taking out non-Americans that's in the book. And it's a tough read in print, you know, and I don't think it, or you tell me, that it wasn't necessarily what they meant, but it just kind of was an indication of kind of what they thought about this mission and what they were going after.

12:56Or how did you read it? Yeah, I read it. And there's a spectrum. So Drew Cukor speaks very eloquently about saving civilian lives and that this is the point of AI. Right. He recruited a very aggressive team of Marines, himself a very aggressive Marine. And some of the people on the team, they weren't only Marines, were very cavalier in their delivery. This is particularly a team that has spent time in war, that has lost people, that feels compelled to deliver better tech, and comes with a fury for the Pentagon bureaucracy that they feel inhibits the ability to develop tools that will save them, their friends, And for some of them, they were thinking of civilians.

13:39But in this case, I did come across examples of people who did say, I want to join this team because I want to reduce the non-American population. And on another occasion, someone said, and I'm afraid there are expletives throughout the book because it's full of Marines, but saying, I want to reduce the expletive. I want to kill people all the expletive time. and I think these are being related to me. It certainly wasn't for public consumption. But people told me those things and they told me because I think they want to explain the breadth of feeling. And I did later say to Drew Cukor, you know, you speak about this very nicely.

14:23And some people on your team speak about this in a very different, harsh, aggressive way. How do you see it? And he said to me, well, when this book comes out, I hope that people see that we are all of America. All American views have gone into creating AI, and the AI comes out will be the product of our decisions. Do you think that's why Drew ended up ultimately being so cooperative with you from being so reticent in the beginning? Was that the message that, like what motivated him to share so much with you? That's always really hard to work out as a journalist, and I never want to ask too much about that once things are happening.

15:06But I do think I did ask him one night, you know, I found out a lot about this story that is not flattering. And I spent a long time saying, I'm not a think tank. I am not the defense tech world. I am going to do this as a reporter. and I found out things that you will not find palatable to be told. Tell me why you're leaning into this story. And the reason he gave me was that he said, one, he believes in journalism and the pursuit of the truth. Two, he's proud of what he did and he would do it all again, exactly the same way. He was investigated. Some of his team behaved, I think by all accounts in the book badly.

15:51And he said he wanted to make sure that the next generation of young officers was aggressive to reform bureaucracy and deliver battlefield advantage. So that was his reason. I think also there was so many, the more I did the reporting, the more I learned there were a lot of clashes within the team. The fault line that I had imagined was between civil society and the military. We'd seen that play out in public. But I began to learn that within the team, there were furious debates about whether they should use Palantir or not, whether they were really developing robust and reliable AI, the ethics of it all.

16:28So it kept unfolding. And he stayed with me through the course. But no regrets, because I was thinking about it as I read it, or, you know, in terms of like Oppenheimer and the, like, regrets of creating things. And I just, but no regret? I don't think he would fall in line with the Oppenheimer idea. Okay. But he is very much a, I end up calling him this sort of figure of, historical figure of future war. Yeah. And other people spoke to me saying there is this tussle between us all to be known as the AI Oppenheimer. And others have claimed it in the newspapers as we started to get generative AI momentum, making these huge claims for how AI would work.

17:13Part of the team wanted me to understand AI fails. There is this big debate. And it's one of the reasons I think I've been able to surface examples where algorithms do go wrong, where AI is not reliable. People are thinking carefully about this and are, even the ones pushing for it, want a public debate about it. There's a quote in the book, and I'm going to be honest, I listened to a podcast and somebody pulled it out. And I'm like, this is really amazing. But AI remains a narrow, faulty tool with considerable limits to its usefulness and reliability that the US military is still discovering.

17:48It's still a work in progress. And research shows that LLMs in particular have a tendency to escalate and agree with you. Now, maybe in a corporate setting, you can deal with that in a war setting, asking an LLM to help you plan a campaign with the risk of it always agreeing with you, shall I strike this target? And it comes up with a reason to justify striking the target. That's not my example. That's an example that military operators have brought to me. So should we bring it into today and the conflict that we're seeing right now? It's not exactly clear what technology is being used and where it exactly is being used and the military wants to hold back on that a little bit.

18:33But what have you been able to find and report about Maven being used in the conflict with Iran? I've reported that Maven is being used. The CENTCOM itself doesn't name the tools, but they've said quite publicly that they're using a variety of advanced AI tools. I've also reported that in Maven's smart system, there is CLAUDE, this LLM that is at the heart of this big fight between the Pentagon and Anthropic. And it's a very interesting fault line, significant fault line to be playing out as the U.S. is using it at the same time that it is helping speed up processes to enable targeting. And CENTCOM said yesterday they have now hit more than 10 ,000 targets through this war, including 1 ,000 in the first 24 hours.

19:14Because of the ability for using MAVEN, right, to identify targets. I think I want to be cautious. I don't want to say it's all because of AI and Maven. Obviously, the US has been able to bring enormous firepower air battles before. But it is clear that even CENTCOM is saying this is speeding up processes. Palantir is loudly calling this, saying, when we look back at this conflict, we will see this as the first large-scale combat. And the words used were driven, enhanced by, and made more productive with the help of AI. There's a moment in your book with literally the chats that went before a drone strike that killed a civilian.

19:54And so what was it like reading that and talking to the drone operator? Which example of this? This was a drone strike that actually killed a civilian? Yes. I mean, things go wrong, I guess is the point. And I guess things go wrong pre-AI. Like, we know that. And I just, I think we're all trying to figure out, though, does AI get over, you know, out of hand when it comes to military battles? And the thing that, you know, that technology companies are worried about, it taking over. And is that how those in the military, those at the Pentagon, see it? Or is there always a human going to be there to kind of work alongside of it?

20:36Not necessarily. A lot of senior military commanders say that U.S. policy is we will always have a human in the loop. That's not true. U.S. policy at the Defense Department is there will be appropriate levels of judgment over the use of force. That implies something different, perhaps supervision. Some people refer to that as on the loop. So the process is happening, selecting targets, executing against targets, and there's a human who can stop it, watch what is happening, but it may be in the future happening too fast. Now, the military commanders who incorrectly use that term, human in the loop, are of course trying to assuage the public opinion, often don't know themselves that language because it's a policy document.

21:21And I think their use of human in the loop is significant because they're the ones actually conducting operations. So that is how they are seeing it and deciding to execute it. But there's a big push to develop autonomous weapons, exactly the kind of weapons that campaigners were worried about from the outset. What is an autonomous weapon? Lots and lots of definitions. In fact, there's a UN body that has spent the last 10 years trying to come up with a definition and disagreeing. But let's say it is a weapons platform that can select, identify, select, and fire at or target an object of its own accord.

22:00So this might be, for example, if you imagine a drone that can fly around on its own without calling back to the internet. It can plot its path to a certain place. It can, using automatic target recognition, assisted by something like computer vision, select a target and then fire at it or fly right into it and explode. The U.S. is trying to develop these. And in the book, I explain a couple of the programs that have really developed and are still in production. One is Goalkeeper. The other is Whiplash. Whiplash is an extraordinary thing that I learned about during the course of this book. It is a jet ski, an autonomous jet ski loaded with weapons.

22:43And the idea is that this could be one part of the U.S. arsenal that if it came to a decision to defend Taiwan, if China ever decided, it's a lot. Well, which takes us back to like one of the things going through this and we're so focused on the Middle East is that this project, right, was started initially with concerns about China. You have to see from 2018 onwards, the overriding effort of the senior folks at the Pentagon is beginning to square up to the chance of conflict with China. Having fought in Afghanistan and Iraq for years, there are these assessments that say China has spent 10 years studying American weak points.

23:22That would be things like the crown jewels of the U.S. forces, aircraft carriers, which are by some analysis framed as sitting ducks. And so the U.S. begins to try and understand how can we be quicker than China, even though our number of forces are fewer. Let's put this in context. The U.S. is, of course, overwhelmingly the best equipped and most moneyed of all the fighting forces and continues to be. But what is China's version of Project Raven? Oh, my goodness. When I get an invite to China to answer that question and give me another 10 years to do it, I will come right back and tell you. But there are interesting research reports that lay out the Chinese ambition.

24:04Interestingly, China starts really talking about its vision for AI after U.S. documents start doing it. So there is an argument to be made that the U.S. started this race. But the U.S. was also animated by fears that China can, once it decides to, deploy, develop and deploy weapons on a much faster, almost twice as fast as the U.S. So although the U.S. was ahead in some ways, it has also always been animated by this fear that it is playing catch-up. Katrina, do you think MAVEN or this type or wherever this goes next and the use of AI throughout warfare might be the ultimate deterrence when it comes to a world war or the next world war?

24:43Could it be the thing? I mean, we've been there before. We thought missiles, right? We think about these things that ultimately keeps everybody in place. Or do you not see it that way? We keep having guests come on and say, you know what, this is our new world order. We're going to continue to see battles coming out and about. You just think over the last four years, right, what we've seen. I'm just curious what you think. All of deterrence theory is changing because you have obviously nuclear weapons, which is supposed to be the thing that stops great powers going to war. And for years now, the U.S.

25:15has observed fighting below that surface, proxy wars, but also this erosion of U.S. military dominance. and whether AI can be the new deterrence is something the people who are trying to create AI are fighting for. Obviously, they frame it not as it is we will win or hopefully we won't have to fight because China will be so afraid of us. And so I've spent a long time trying to work out how much might be bluster and positioning to try and deter a war and how much these tools are actually reliable. You know, it gets me thinking about somebody who has done the reporting you've done and synthesized it into a book such as this.

25:57Do you feel, based on what you know and what you've reported and what you've found, do you feel like the world is safer as a result of this technology? Please say yes. Please say yes. No, I'm just kidding. I think back to a book I read when I was much younger that was written by a Polish banker named Ivan Block in 1899. And the English translation of it was, is war now impossible? And the argument he explored was that because the world was producing at wild scales rifles, because weapons had got better and better and better, the claim was made that no one would dare go to war. And he already argued the irony of this, that better weapons would not make for quick wars.

26:45they would make for long wars, for stalemate, for terrible human horrors. And of course, 15 years later, there is World War I. His paper is one of the reasons that I wanted to write this book, because we are at a moment where everyone is claiming bigger and better weapons will keep us safe. Of course, America is the only country ever to have used the nuclear bomb. And the argument is still made that America had to use it. And the world is really divided because there are just as many, perhaps more people who argue it was a terrible decision and people have regretted it. Einstein, notably. There's a very interesting paper where Alex Karp in an opinion piece, of Palantir, yes, the chief executive, he referenced Einstein.

27:35But in my book, I go on to say he omitted to say that Einstein said that pushing for the bomb was, in his view, the biggest mistake of his life. We're going to leave it there. Thank you so much. Thank you so much. The book, as we said, Project Maven, a Marine colonel, his team, and the dawn of AI warfare. This is the era we are living in. Incredible reporting. Katrina Manson, you know we adore. Bloomberg News reporter covering tech and national security. It is a must read. It is out. And check it out. This is Peter.