Why Iran is Attacking Data Centers w/ Sam Biddle

Paris Marx is joined by Sam Biddle to discuss what it means for data centers to become targets in a war, and how Silicon Valley is aiding the US war against Iran.

Sam Biddle is a technology journalist at The Intercept.

Tech Won’t Save Us offers a critical perspective on tech, its worldview, and wider society with the goal of inspiring people to demand better tech and a better world. Support the show on Patreon.

Take your personal data back with Incogni! Use code SAVEUS at the link below and get 60% off an annual plan: https://incogni.com/saveus

The podcast is made in partnership with The Nation. Production is by Kyla Hewson.

Also mentioned in this episode:

• Send your questions to mailbag [at] techwontsave [dot] us!
• Sam wrote about Iran’s attacks on data centers and its legality.
• Here is Sam’s most recent piece about Palantir and NYC public hospitals.
• The Intercept is also covering the role of social media in the US-Iranian war.

Transcript

SPEAKER_02 0:00

And so to switch from ChatGBT to Claude, I think it accomplished nothing and signaled very little, but it gave people the the at least the approximation of um some sort of political act. But I mean it's it's like it's like switching your business from Lockheed to Raytheon, right? Like I I don't there's not that much difference at the end of the day.

SPEAKER_01 1:00

And now, you know, the show has reached, I don't know, tens of thousands of listeners every single week. Uh who knows how many that is cumulatively? It's in the millions. Uh I should have looked up the numbers, I guess. Um, I've had more than 320 episodes. That means so many interviews with fantastic, insightful people to help inform you about the tech industry, the impacts of this industry and its products on society, and how we might think about doing something better, right? How we might think about challenging this and trying to imagine what a better world could look like and what better technology could look like as well. I feel like tech won't save us has achieved a lot over the years. And I have had so many positive messages from people about the show and about what it's meant to them and about how it has allowed them to share these episodes with people to try to get more people to think about technology in the way that we do. And so thank you so much for listening for all of this time. Uh, it's it's, you know, still hard to believe that it has been six years of making this show, um, but it wouldn't be possible without all of you, the fantastic listeners. And I know that this year, you know, the past year, I feel like has been a bit slower for Tech Won't Save Us than it usually would. We weren't able to do a special dedicated series in the fall, which was my plan, uh, because I just ended up being so busy working on my next book, which comes out in October. Stay tuned. Uh, there will be more information on that soon. But luckily, the work on the book is finally winding down. And that means more time, you know, more attention for Tech Won't Save Us. And there is a lot I want to do with the show in 2026, some things that we've already started to do, you know, some of the interviews that I've had already this year to give an indication of the type of things that I want to do with the show this year. But there's more to that as well. And I need your support to do it. So where the show is turning six this month, I'm hoping that we can get a hundred new supporters over on patreon.com slash techwon'save us. If you are not a supporter of the show already and you really enjoy these interviews, you feel like you learn a lot from them, I would ask you to consider, you know, putting the show on pause and heading over uh to become a supporter now to help support this work so I can continue to do it and so that we can, you know, increase our ambition and our scope heading into uh, you know, our sixth year uh of doing this show. I am working on a new series now that I'm tentatively calling it the Internet Deception, but maybe I'll change that by the time it actually comes out. And we're aiming to have that out in May. And this will really be a series about the privatization of the internet, you know, how it set us on the path that we have arrived at now, what we misunderstood about it at the time, you know, with some of the narratives around what the internet was and what it could be, um, and how, you know, recognizing that there was a problem with how we misunderstood what happened with the internet back then and how it has developed since, how that should force us to change how we see the internet and how we approach the internet in the present, right? Because if the history is wrong, right, if if the foundations of what we have built our, you know, understandings on and our narratives on around the internet was wrong, then that forces us to change some other things as well. And that means approaching the policy through a different lens, too. And so I hope that is going to be very informative for people. Uh, as I said, we're planning to get that out in May. And people who are Patreon supporters will get full access to the series before anybody else. And then after the series finishes airing, they will also get access to all of the interviews that I did for the series as premium episode. So if you want to get access to that, it's a great time to become a supporter to, you know, support us to make the series, but then to get access to that additional content that is going to come out around it. Uh, I will also be recording a mailbag episode soon, which I know I said I was going to record like a couple months ago, but then edits on the book just became a bit too much. And unfortunately that had to get uh pushed off. So if you do want to, you know, contribute uh a question before I record it, you can certainly send one to mailbag at techwon'save.us. Um, and I will get that. Uh, we'll put it in the show notes if you if you want to find the email address there. And so then, you know, you can ask a question and I will answer it. Uh, and we're certainly hoping to do more kind of premium content through this year as well. So stay tuned for more information about that. Of course, the show has also gone video. Uh, you know, if you didn't notice last week, we now are doing video episodes on YouTube and Spotify. Um, I believe Apple is going to offer that at some point as well. Uh, you know, I wasn't super enthused about video, but that is the way that things are going. And so Tech Won't Save Us is doing it as well. So if you really want to see me uh, you know, interviewing someone else, you can you can do that through the video platforms. But if you like audio, you can keep listening on audio too. So that is a lot, I know, but you know, the key points are thank you for supporting the show, for listening to the show. And if you're able to do so, I would ask you to go over to patreon.com slash tech won't save us, become a supporter today, so you can support the great work that we continue to do and hopefully that you, you know, enjoy every single week, and so that we can continue to do the types of things that we want to do this year and into the following years with your support, because listener support is essential to what I do with Tech Won't Save Us and to give you, you know, this quality of critical analysis of the tech industry. And of course, if you do become a supporter, you will get access to our backlog of premium and ad-free episodes. You will be able to join our Discord and you know, join in the discussions that happen there. You'll be able to get stickers if you support at a certain level or above. So thanks again. And go over to patreon.com slash tech won't save us to help us hit our goal of 100 new supporters this month to celebrate the sixth anniversary of TechWon't Save Us. Now, with that said, onto this week's episode. My guest is Sam Biddle. He's a technology journalist at The Intercept. And if you listen to last week's episode with Spencer Ackerman, we're where we talked a lot about kind of the history of what got us to this point where you know the United States and Israel are at war with Iran. This is really a follow-up to that conversation where we dig much more into the consequences of this, right? With a particular focus on the bombing of uh Amazon data centers in Bahrain and the United Arab Emirates and what that really means, right? What it means now that data centers can be targets in a war, recognizing that, you know, obviously AI tools and the cloud are used for targeting and other military applications. So it seems pretty natural that things would move in that direction. But even still, it seems pretty shocking, right, to imagine a data center actually getting bombed in a war and what that means for the future of these infrastructures, you know, how we see them, but also what it means for the collaboration between Silicon Valley and the military that we have been seeing escalate for quite some time now, um, but now has reached this point where it's quite undeniable with the Iran War and, you know, with the other wars that we have been seeing in recent years, not to mention the genocide in Gaza. But really, I think that this is a fantastic conversation with Sam to get into the wider repercussions of this, the wider ramifications of what this all means. Um, and Sam is just the perfect, most insightful guest uh to basically discuss this with us. So I think you're really going to like this episode. I think it's a great way to start off this month of the show's sixth anniversary. And so I think you're really going to like it. If you do, uh, you know, again, head over to patreon.com slash tech won't save us, become a supporter this month to help us hit our goal of a hundred new supporters. I would really appreciate it. And it helps me to continue doing these great interviews every week that provide you the critical perspective that you want and that people need about the tech industry. So thanks so much and enjoy this week's conversation. Sam, welcome back to Tech Won't Save Us. Thanks so much. It's great to be here. Absolutely. It's always great to talk to you. You know, I'm always reading your stuff at the Intercept. Um, and it's always, you know, fantastic to talk to you about what you have been working on. And, you know, of course, we're all very focused on this war between uh, you know, the US and Israel and Iran at the moment. Um, you know, obviously it's you know in the news every day. It's hard to know exactly what's happening. And just to be clear, we're recording this on March 25th. So there will surely be changes and developments by the time that this um gets published. But, you know, obviously now this war has been going on for a number of weeks. Um and before we dig into, you know, like the core topic that I wanted to discuss with you in this episode, I was just wondering if there's, you know, anything that really stands out to you as you are watching this war progress, you know, whether that's specifically on the use of tech or what we've seen technologically, or, you know, just more generally about what this says about kind of like geopolitics in the world that we're living in right now. So yeah, what stands out to you about uh what we're seeing?

SPEAKER_02 9:37

I mean, I think what's been most striking to me is as you mentioned, how little we know um about a war that is so thoroughly documented, both by news media and people on the ground. Uh and it's been weeks, as as you also noted. And you know, the the answer to the question of how is the war going for the US and Israel is uh I mean completely obscure. Um I mean, certainly not well in that it's still going. Um I I think it's pretty clear this was not the uh well, I don't want to say the plan, because I don't think there really was much of a plan, but this was not the hope of the administration.

SPEAKER_01 10:17

It was certainly not a uh Venezuela 2.0 situation.

SPEAKER_02 10:20

No, they didn't, you know, sweep in with shock and awe and then get right back out and um you know the world uh most of the world sort of moves on as was the case with Venezuela. I mean to point out the dishonesty of the Trump administration, what I mean, what's even the point of this point, but it the the effect of the uh president and commander in chief, ostensibly the person in charge of the entire thing, being I you know, probably literally the least credible uh source on that war. I mean, not that a president is ever going to tell hard truths about a war, um, but uh uh it's just a complete um unreality. You know, it it it's it it does a number on you, I think. And you know, again, not you know, to say nothing of the people on the ground in Iran who are suffering at the brunt of all of this uh violence and destruction. But um it's very, very hard to be an informed citizen right now. Um I I mean and and that that's it's in a way that I think uh at least I have not seen before. I mean, I was um I'm old enough to remember very well post-9-11 uh wars in the Middle East. And um, as much as those went down the rabbit hole of um, you know, dishonesty and propaganda and so forth, I don't recall it being quite this muddled.

SPEAKER_01 11:37

Yeah, it it's also wild, you know, speaking of like the information environment, to see the way that generative AI has become like a key part of the information war on it, you know, when it comes to these like modern wars and modern military activities that the United States and these other countries are involved in. Obviously, you have, you know, the White House itself kind of producing these um, you know, AI generated, you know, basically propaganda, as you say, in order to justify what it's doing and make it look like it's doing things very well. But then you also see Iran producing its own kind of like, you know, AI generated counter-propaganda. It's I don't I don't know, it's terrible.

SPEAKER_02 12:16

I mean, I I I've also been I've been struck by just how like low effort a lot of the American propaganda is. I mean, they seem there's someone at the Pentagon who clearly, or maybe Hegseth, I don't know, who loves these just like highlight real kind of videos of like basic explosions set to like shitty music, uh in you know, with scenes from like Halo and stuff spliced in. Like it's the kind of thing that like I would have probably thought was really cool when I was, you know, like a ninth grader, you know, like like a child. Um but this you know, sort of like celebration of violence, it isn't even really making an argument. It's just like it I mean the the over the the the takeaways or you know the ethos of just like fuck yeah, you know, but there's no like there it's not even really propagandizing, it's just um it's it's like it's like a marketing clip. But I think the most effective propaganda has been the uh not AI generated um you know footage of Iranian uh missiles streaming over the sky and Tel Aviv or you know elsewhere. Um I mean, I I think that you cannot manufacture or fake or generate something that's as compelling as you know real footage filmed by some person of uh you know this this country we were told would fold very quickly, continuing to fight and drag this war out. I mean, I think that's probably up there in terms of um effective messaging.

SPEAKER_01 13:44

Yeah, absolutely. I I think you're completely right, you know, and it has been kind of shocking to see that the degree to which this has been able to continue playing out, you know, and how how resilient the Iranians have been, which, you know, maybe on the one hand shouldn't be a surprise. Um, but then when you consider especially the way that the United States and Israel kind of present themselves and, you know, uh how they often kind of you know are able to bomb these places and get them to, you know, kind of get in line with what they want and then kind of move on, or at least that's the doctrine that we have seen them want to want to present in in recent years. Um yeah, it is it is kind of wild to see the way that they seem to have really miscalculated um in the case of Iran and really not understood the degree to which you know the the the state could really um you know keep active and and you know keep them at bay in a sense.

SPEAKER_02 14:35

Yeah, well, and you know, when and when Trump, you know, you have the president of the United States saying uh, I think on multiple occasions, including right after the war started, the war is over, that it's been won, that we've crushed the regime, etc. And you know, and that's obviously not true. But the I think the most effective counterargument to that lie is, again, uh footage of a ballistic missile uh you know and entering the atmosphere over a city. I mean, that is that that is as I think effective a uh refutation as you could get of um any kind of claim that this war is has been is is even in the neighborhood of over.

SPEAKER_01 15:12

Yeah, you know, at least he didn't appear on uh an aircraft carrier with a big uh mission accomplished banner. Yeah, not yet. Maybe we'll get there. Maybe we'll get there. But you know, you you mentioned the drones and and the bombs and everything like that, right? And there are many different aspects of this war that we could talk about, um, you know, and that are very much related to your reporting. But there was one element of this that obviously really stood out to me as someone who is very interested in the tech dimension, but you know, has also been following data centers and that kind of story very closely recently. And that was that early on in this war, Iran actually targeted Amazon data centers in the UAE and Bahrain. Can you tell us a bit about what happened there?

SPEAKER_02 15:54

Yeah, so it was in the first day or two. Uh I believe the first strike was uh March 1st. I believe three, if I'm recalling correctly, three Amazon AWS cloud facilities were um hit by uh Iranian drones uh yeah, in in Bahrain and uh UAE. And um it was the damage was sufficient uh to majorly disrupt these cloud regions. And to the extent that Amazon actually told customers relocate your uh cloud workloads to another region elsewhere in the world, which is you know uh not the whole point of the cloud is that it's redundant and resilient and uh you don't have to think about things like geography, right? Like it's supposed to just work. Yeah, I mean, Amazon uh said that there were fires at the facilities, you know, there's obviously fire control mechanisms built in that then flooded it, you know, these these things with water. If you're talking about tightly, tightly packed uh uh uh computers in a giant uh warehouse, fire and water are um uh a pretty major problem. And and so, yeah, I mean it it it disrupted in a pretty major way the one of the flagship technologies of one of the flagship companies of the uh American economy. Whether or not there was a a military um objective accomplished there, um, you know, you want to talk about propaganda, being able to even temporarily shut down uh Amazon cloud stuff is is a pretty big deal.

SPEAKER_01 17:20

No, definitely, right? And especially where data centers have very much been in the public focus, especially in the past few years, right? You know, as the generative AI boom has resulted in so many more of these infrastructures being built, and we're seeing this growing, you know, public backlash to uh, you know, the construction of these facilities in so many communities around the world, um, and certainly across the United States as well. Seeing these targeted in this way was was really, I don't know, again, maybe it shouldn't be surprising, but but it was in a sense, because you know, for so long you didn't think about the infrastructure that that powered the cloud. Has Iran talked about what its motivation was for targeting these data centers specifically?

SPEAKER_02 18:00

So there was a there was a messaging out of um an Iranian sort of state-affiliated media outlet that, you know, I I think you can sort of take as, you know, not the words of the state itself, but reflecting um uh that that rationale. And they just said they wanted to uh highlight the usage of um American cloud providers and data centers as military infrastructure. Um I I think that the I mean this is me speculating a little bit, but I think maybe it was as much to draw attention to the fact that data centers uh have military use as it was to actually um disrupt uh that use. Um I mean, I I I very much doubt that um either the US or Israeli militaries were directly using the data centers in question, but they are certainly using other Amazon data centers um uh you know elsewhere and and in fact very close by within Israel.

SPEAKER_01 18:59

Yeah, absolutely. And and there has been a lot of attention even on that piece in in recent years as well, right? Especially as we've been seeing what Israel has been doing um in Gaza, the way it's been using technology there, and the way the cloud has uh been part of that. Can you talk to us a bit about how you know these these cloud computing centers and you know AI more generally are used for military purposes and how we're seeing them being used um in Iran specifically or or in wars uh throughout the Middle East?

SPEAKER_02 19:27

Sure. So I mean I think that the most relevant um and probably at this point best documented in terms of public reporting uh use case is target generation. Um so we've seen in in Gaza uh over the past few years to devastating effect the um the the the very fast pace and indiscriminate nature of uh bombing there. And if you are a uh military like the IDF that has a basically insatiable appetite for things to destroy in the Gaza Strip, you don't just want to carpet bomb it from top to bottom for I mean essentially PR reasons. Um I think that generative AI allows you to very rapidly uh create a list of people and places that you can um attack, at least have some bureaucratic plausibility um according to this, you know, according to the computer, right? I mean, not according to any anyone's actual judgment. But it gives you an official-looking list of things to blow up. And I think that if you want to blow shit up, that's what you really want is uh a computer that provides the sheen of intelligence and rationality, right? In in you know, in in Gaza, um you know it's been uh uh I think again very well reported by outlets like um The Guardian and and 972 magazine in particular, um you know, uh the use of uh generative AI tools to um you know to find the to suggest for bombing the homes of you know alleged militants, uh, you know, whatever that means, uh however many degrees removed from any actual militancy. And I think that that uh is maybe one of the most concerning and and quite literally dangerous use cases for AI to just generate uh a list um and you know what happens to the people and places on that list.

SPEAKER_01 21:22

Yeah. And I I feel like you're getting to the two different angles of this, right? On the one hand, it is the use of the technology to generate these targets that can then be bombed, right? And and you can generate a really large list in a very short amount of time. And you know, whether or not a human checks that or not, uh I think it's fair to say it's often not being checked. Um, but it, you know, it gives you a long list of of places that you can target that you know the computer thinks are associated with you know the the enemy or or whatnot, right? But then on the other piece is like, okay, so there's the the generation of all these targets, but then because Because the computer has generated them, it's like it starts to reduce the accountability for the humans that are then actually, you know, pulling the trigger on where to drop these bombs, right? And and yeah.

SPEAKER_02 22:11

I mean, not that there's very much accountability when it comes to air, you know, air war generally. Um, I mean, you know, pre-AI, we were certainly still we, I mean, you know, the the the US where I live was certainly killing civilians based on targeting that had been done by um uh uh humans. There's uh scant accountability when it comes to that. Uh but yeah, I mean, even on paper where you're supposed to have a system of legal accountability or congressional oversight onto these things, the computer gives you a list of you know a hundred things to blow up and you know 20 of them were wrong, uh, whose fault is it? I mean, uh the the computer doesn't have personhood, um, it doesn't have moral agency, it doesn't have anything. I mean, it's just a computer, right? So you can only but the it then you know who you're then supposed to blame, I think becomes an open question, right? I mean, is the person is it whoever trained the model? Is it whoever queried the model? Is it whoever signed off on it? I mean, I I uh in a world where we where there is very, very, very little accountability, I think this further suppresses the possibility to hold anyone accountable. Um, I mean, I you know what I think about a lot is like, so one use case that I could see being potentially useful in my own work with when it comes to an LLM is like summarizing documents, right? Which I think is uh a major use case in the military and intelligence worlds as well, right? Like you have a huge pile of paperwork, dossiers, old satellite maps, you know, uh, and you want to feed it all into the machine and get a summary and then, you know, uh uh plan operations based on that. The thing that's always stopped me from using it in my own life is if there's even like a 5% chance that, or you know, if there's if like 5% of the summary is going to be wrong, right? And I'm pulling that number out of nowhere, but we know based on how these things work that there will there's a good chance that some of it will be incorrect or hallucinated or whatever. Then if I don't want to embarrass myself professionally, I gotta go through and read the whole thing anyway to check what was wrong. So I might as well just read it myself, right? Like it's it's not actually saving any time. Um, and I think about that a lot in terms of uh, I mean, you you brought up the question of like, is anyone, you know, signing off on this stuff or going through it to verify it. I can imagine that in the military and intelligence world, you have the same issue where like if we're going back through to double check the computer's work, it sort of defeats the purpose of using the computer. So I I I would be very suspicious of how much uh how much they are checking uh, you know, Claude's homework uh on this kind of thing.

SPEAKER_01 24:47

Yeah, I I know in the case of like the reporting that 972 magazine was doing on what was happening in Israel, there was a suggestion that not very much of this was being checked by humans after it was being generated, right? And just to pull up another example based on what you're saying, obviously one of the first major stories to come out of uh the US-Israeli war on Iran was, you know, when they bombed this girls' school, right? Um, you know, in Iran and more than 170 people were killed, most of them schoolgirls. And we had a number of stories after that seeking to justify or understand why that happened. And one of those stories was it's possible that the AI target generation system uh, you know, misread a document or that this facility was previously associated with the IRGC or something like that, and has since not been associated and thought it was still. And it and it's like, you know, it just adds to this like complexity and again, like the lack of accountability around it, right? It's like, oh, well, maybe the computer just got it wrong and we're so sorry that this happened, like blah, blah, blah, right?

SPEAKER_02 25:46

Yeah. I mean, I so I I have not seen anything that I all of the talk I've seen around the use of AI in that in that specific strike on the school has been speculative, right? I I have not seen any kind of confirmation. I'm not saying any of that is concrete, of course. Um and yeah, it's completely plausible, right? I do worry that maybe the like jump to speculate that basically the jump to blame AI for something like that obscures the fact that humans are perfectly capable of bombing schools on their own, um, or weddings or hospitals, right? Like you do not need an advanced technology to kill civilians, even by accident, right? Like you even let's let's give the the US I mean I don't often say this, let's give the US military the benefit of the doubt here, right? And and this maybe this was genuinely uh an accident. I I would argue a you know criminally negligent accident, but it it these things do happen based on recklessness and indifference to human life and um carelessness. Historically, there are you know many cases predating uh uh generative AI or LLMs or whatever. So, you know, I I I get the uh appeal of connecting this new technology to something horrific in in current events like that. But um yeah, like I like just like I said, humans are perfectly capable of um being indifferent to uh you know the facts on the ground and killing people over it. Um so I I maybe it was based on you know an AI mistake, but um yeah, I I I think that requires more uh reporting and scrutiny.

SPEAKER_01 27:26

Absolutely. And you know, just to add to what you're saying, it's like we see plenty of cases of that right now, whether that is the bombs that are being dropped in Tehran on oil facilities, like you know, the infrastructure that that regular people depend on, but also what we've been seeing over the past number of years in Gaza, in Lebanon, in Ukraine, you know, we can see that humans are gonna do a lot of really terrible things to one another if they have the tools to do so, um, you know, and and this kind of animosity is created and and whatnot. So we don't need to say, we don't need to pin this all on the technology uh when we know that humans are gonna do this, unfortunately, if they have the ability. Yeah.

SPEAKER_02 28:05

Yeah. I mean, I so the I the there was a there was a report of the Washington Post in the in the first several days of the war um about the use of Claude of, you know, anthropics LLM that was being accessed through uh Palantir's Maven Smart System targeting platform. And uh just revealing that uh, you know, the the post reported that Claude was being used to plan um airstrikes in Iran. And what stood out to me was not you know more than just the the the the sheer fact of using Claude for this purpose, but the the post report uh said that they were using it to speed I mean I'm I'm paraphrasing here, but to basically speed up the uh the targeting uh and execution process of these strikes. And anytime you see that, I think that alarm bell should go off. Um because uh the the I I think you in in uh Hegzeth and Trump, you see a clear desire to just blow the shit out of Iran as much as possible, as fast as possible. And you know, and anytime you're accelerating uh killing, especially aerial uh warfare, airstrikes, I think that really, really, really cranks up the likelihood of killing innocent people. Because uh because you were not and and this again, there there are many examples of that pre-AI. I mean, like in the war on terror, there you know, I I I um there's been some excellent uh reporting around how uh JSOC really wanted to just I I think the you know the the term they use is like uh pick up the the operational cadence, right? They just wanted and and that was uh you know it was pre-AI, but it was very much computer enabled, you know, just like let's you know they they wanted to just aggregate all the data, let's see, pull out names, pull out places, and go kill people. And the the it was done the speed was the name of the game. And so I think when when you have a desire to just uh not just execute a war but execute it rapidly, that is extremely dangerous. And and I do think that you know, whether AI was used in that school strike or not, a major uh function it provides in military is being able to you know hit the gas on killing. And you know, and and that's something people should be extremely disturbed by.

SPEAKER_01 30:17

Definitely. It it brings a whole new spin to move fast and break things, you know, just not over. No, absolutely, yeah. So long associated with them.

SPEAKER_02 30:24

Right, yeah, right. Move fast and kill people is is I think um you know the the the ethos of um AI augmented uh warfare. Um you know, and it I I I do also um you know, I also don't want to say that like it that this school strike, I'm not trying to say that it like wasn't uh done with AI. It's entirely possible. I just don't know. And I, you know, I don't want to say one way or the other. But um certainly it's a question that be that uh that people should be putting to both Anthropic and the Pentagon. I mean, I think we're owed an answer on that exact question.

SPEAKER_01 30:56

Yeah, I no, I completely agree. And you know, I I I wouldn't give them too much uh what benefit of the doubt on on that strike and try to try to explain it away with AI, especially when you know we've had the reporting now that shows it's a double tap strike as well. So it's like, I don't know, I think they kind of knew what was happening if if they're going that deep with it, right?

SPEAKER_02 31:16

It's inexcusable no matter what. Like there's no possible ration ra rationalization for kill for uh hitting a school. Doesn't matter what used to be there or what was next to it, right? So I mean I mean I know that you agree with that, but uh you know, the the the AI or no AI, it's it's a crime. Um so uh you know I think that's what's most important.

SPEAKER_01 31:35

Yeah, and maybe just to be clear to listeners, double tap strike is when they strike it once, and then once people kind of gather around to try to, you know, uh rescue people or whatnot, they they hit it a second time, right?

SPEAKER_02 31:46

Um it's a it's a tactic that has been used extensively uh by the Israeli uh Air Force.

SPEAKER_01 31:52

Yep, absolutely. If you've ever thought, wow, that scam email feels weirdly personalized, that's not intuition. That's the data economy working exactly as intended. That's why I want to talk about incogni. When your home address, phone number, email, and other personal details are sitting in searchable databases, it increases your exposure to phishing attacks, identity theft, and financial fraud. This isn't just about annoying spam, it's about real risk. Incogni works to remove you broadly from the internet. They identify where your data appears and send legally binding removal requests under laws like GDPR and CCPA, follow up until deletion is confirmed and continuously monitor so it doesn't quietly reappear later. And what I really like is their custom removals feature. If your information shows up at a specific site outside their automated system, their team will manually pursue that removal too. That's included with the unlimited and family unlimited plans. And honestly, if you want the most comprehensive protection, especially if you're thinking about your household, that's the route I'd recommend. This is about shrinking your digital footprint and lowering your exposure to scams and identity theft and not pretending the problem doesn't exist. Right now you can get 60% off an annual incogni plan. Go to incogni.com slash save us and use code save us. You can't dismantle surveillance capitalism overnight, but you can make yourself a less convenient data point. Again, that's incogni.com slash save us, code save us for 60% off today. You mentioned Claude there, and you know, I there's a whole conversation that we could have about what happened during the war on terror and the developments in in technology then. And I think we might even have a previous episode on that. I'm forgetting that there's been so many episodes now that I mean I gotta listen to this.

unknown 33:30

Yeah.

SPEAKER_01 33:30

Um, but you know, like even Palantir like kind of comes out of that moment and things that were happening, right? As you've mentioned. But one of the big stories around this war in particular has been the Pentagon's desire to have a generative AI partner um, you know, for military purposes. And before the Iran War started, we basically had all this reporting on whether Anthropic, you know, which makes the Claude chatbot LLM was going to basically sign this deal with the Pentagon. And, you know, eventually they decided that there was a piece of that contract that they weren't gonna go for, and OpenAI signed a contract instead. And then like, I think it was like the day after they signed the contract, they started bombing Iran or something. Like it was a very bad look for Sam Altman and OpenAI, and they had plenty of you know things to be feeling bad about.

SPEAKER_02 34:22

I think it was literally later that day. Uh you're probably right.

SPEAKER_01 34:26

Yeah. Like it was, it was very not long after, right? But then, of course, we have these stories, and and I will say that prompted a whole campaign, like, you know, drop open AI, start using Anthropic and Claude instead, like a consumer campaign, uh, quit GPT, I think it was called. Um, but then of course, we started getting these stories about how the Pentagon was very reliant on Claude actually for this war, and as you're saying, for the target generation and things like that. So what are we seeing with the Pentagon's relationship to these AI companies and its use of different services and things like that in order to prosecute this war, but also just in general with the types of things that the military is doing?

unknown 35:07

Yes.

SPEAKER_02 35:07

I mean, I something that really stood out to me with the um the clawed GPT, you know, dust up and rivalry and like you said, the sort of consumer response of like, you know, boycott open AI, let's all use Claude. I think it sort of created this narrative around anthropic that it's like code pink or something, that the this was like people who were gonna stand up to the Pentagon, which is just uh absurd. I mean, they are they are uh a uh military contractor and a very eager one. I mean, the the company put out a statement in in the days following uh the uh I think it was right before the lawsuit they filed against the um uh uh about their um designation as a supply chain risk. But they put out a statement saying, and I'm I'm paraphrasing here again, it was something like we have more in common with the Department of War than uh differences. I mean, like uh and that that's in the midst of, you know, arguably an illegal war of aggression uh being perpetrated against Iran, in which many, many innocent people are being uh injured, maimed, or killed. So this is not like the peacek AI lab, right? Like they have this is a disagreement over some very, very, very narrow specific use cases. The thing that that uh anthropic says they did not want to engage in, namely uh fully autonomous lethal weapon systems and uh mass domestic surveillance. Um, which, yeah, I mean, I I I I it suppose it's good for society that there were there was some uh conscientious objection to those uh applications, which clearly are not held by uh OpenAI, but um that still leaves a vast universe of of death and destruction that uh anthropic is willing to assist in, right? You know, they're they're not saying this, but it is it you know, implicitly they are a company that is okay helping a military that is killing civilians, right? Um, so uh I yeah, I I I thought that was very I thought that was very funny. Um I mean, maybe funny is not the right word, but you know, I think people are so I think people feel really helpless, right? And so to switch from Chat GBT to Claude, I think that accomplished nothing and signaled very little, but it gave people the the at least the the approximation of um some sort of political act. But I mean it's it's like it's like switching your business from Lockheed to Raytheon, right? Like I I don't there's not that much difference at the end of the day. Uh and you know, I I wouldn't I would not be surprised if this lawsuit uh prevails and um uh anthropics lawsuit against the Pentagon, to be clear, um, if they prevail on that and are right back where they started. Um I mean, this again, the most important thing is uh this is a company that very much wants to be a defense contract, to be a military contractor. That's who they are. Um it's it's a real trip. I I suggest I recommend anyone who wants to feel really depressed and fucked up for a little while to read through Anthropics. They have a like uh clawed constitution and like a policy document of things you can't do. And there's so much talk in there about how important it is to anthropic to avoid doing harm. And like it's you know, it's it's absurd. This is their their flagship product is being used to plant airstrikes. I mean, uh to then talk in the same breath about not doing harm is just again, I mean, it's it's it would be funny if it weren't so uh so so perverse, you know.

SPEAKER_01 38:40

Totally, because like the harm they're talking about is like not creating the uh what the the AGI that might kill us or something rather than like you know the real harm that is being caused day in, day out by the use of its tool, right?

SPEAKER_02 38:54

That's a whole other that's a whole I'm sure I I have I think you've probably done an episode of this, but that's a whole other conversation about like what safety means in AI researcher world, right? And it is very much not about the common sense understanding of safety and harm, right? Like uh yeah, I think you're probably right that if you ask them to defend that, they would say, Oh, well, yeah, the harm we're talking about is uh the internet becomes sentient and starts building, you know, death robots or releases you know, sarin gas into I mean, I right, I mean like like far future sort of science fiction scenarios. Uh meanwhile, they are actively harming people probably every day across Iran.

SPEAKER_01 39:32

Totally. And it it's been it's been fascinating too to see like the evolution in that too, you know, because a number of years ago, there was an explicit prohibition within you know open AI's rules or whatnot, uh, you know, against military uses of its AI applications. And now it's like they're the ones jumping on uh the Pentagon contract to try to try to do it, to try to get you know more use of this tool, to try to get a new uh you know big contract, big partner. Um, and even though Claude has has lost this contract or you know, was they were not willing to agree to like one specific uh you know type of thing that they didn't like in the agreement, um, as plenty of reporting has showed, and as you've mentioned, it is being actively used uh to you know identify targets in Iran. And the reporting I've seen suggests that as much as the Pentagon might not want to use like the woke claude or whatever, that it's pretty impossible to get off of at this point because it's so like deeply rooted into the systems that they're using.

SPEAKER_02 40:31

I think that I think a lot of people depending on probably really love woke claude, right? I mean, as evidenced by the fact that they're using woke claude. I mean, this this was sort of a weird spat between uh, you know, Heg Heg Seth and his uh deputies. And you know, I I think frankly, they just didn't like being told no, or even have the suggestion of a no coming from a uh military contractor, which is I, you know, I mean the hypothetical that I've seen brought up a lot from people who are defending the Pentagon's uh course of action here was well, can you imagine if like Raytheon told the Pentagon how, like when they could and couldn't drop bombs, which like, yeah, fair enough, is an absurd arrangement. And and of course Raytheon would do no such thing. But it's it's almost like taken what's missing from this whole conversation uh is like anthropic doesn't need to be a military contractor, right? Like they could be a they could be a consumer or research, or I mean they could service every other part of the economy, um, and and and even government, right? Like no one has forced them or coerced them to pursue Pentagon money. Um so I think that there is very much a sense in which they want to have it both ways, but I I think the most important thing is that the purported and self-professed values of these companies are essentially meaningless. Like you, you, you know, Google had a, you know, you can have a document that says do no harm or don't be evil, but as soon as you want to do harm or be evil, you just remove it from the document. Like they are the and and like the fact that they're often called like constant, like clause is called the constitution, which like amending the US Constitution or any nation's constitution is like a really long process, difficult process. I mean, in the US, it's nearly impossible by design. The document is supposed to be not immutable, but very, very, very hard to change, is which is what gives it power. Um, whereas uh tech company policy documents are constantly being changed, they don't mean anything, they they are shifted according to the whims of uh the you know executive leadership. So I I mean I I really do think that you should anytime you're looking at one of these policy documents, sort of acceptable use policy, and this this goes for outside of AI uh stuff too, but they are not worth the you know pixels they are printed on. They they are they are um, I think window dressing and sort of a public relations instrument and and nothing else. You know, if they want to change the rules, they'll just change them, which is exactly what open AI did. They removed the language from their permissible use policy, forbidding warfare, and then pursued all those contracts. So, you know, it's it's very it's silly. Oh, it definitely is. I mean, it's not silly, it's it's deadly serious, but it's it's the documents themselves are silly.

SPEAKER_01 43:17

Yeah, yeah, I get you. But I I want to kind of broaden out what you're saying, right? Because we're talking about the AI companies specifically, you know, uh Anthropic, OpenAI, but we started this conversation talking about an Amazon data center, right? Um, and and the targeting of this cloud computing infrastructure. Um, and it's it's very clear that you know you you were saying Anthropic could do everything but working with the military, right? And they have chosen not to do that and to pursue these contracts. And I feel like we're seeing that a lot from these tech companies going harder and harder at uh, you know, working with the military, working with the Pentagon, seeking out these contracts, um, whether that's for profit reasons or for broader political reasons. What do we see in the broader kind of tech? tech industry turn toward the military and how does that contribute then to say an Amazon data center then being seen as a target in war?

SPEAKER_02 44:09

Yeah I mean I think the the inflection point was um Project Maven which was the uh uh aborted uh Google uh uh deal to do um uh aerial warfare targeting uh basically to aggregate and analyze uh data that could then be used to plan and execute airstrikes um there was a pretty significant employee revolt over that um there was a lot of news media scrutiny there were protests and Google walked away from it and said look you know uh enough people within the company have said they do not want this to be the kind of the they don't want this to be the way uh we make money so we're not gonna do it and there did seem like a time when uh opposition to militarism within Silicon Valley was sort of the prevailing uh mood so Project Maven was uh launched in 2017 by 2018 Google has abandoned it so um you know pretty rapid turnaround and again I mean the the the org the organizing by workers at uh at Google um was very effective. I mean it they they they won. You know that is that era is very much over. I think that companies like Amazon have real and Google have realized that they can just plow forward, you know, fire employees who protest get the money. And also I think the sort of worker sentiment has changed. I think you see more people joining these companies that say no I do want to help the military you know like I I'm okay with building um helping build bombs and rockets and and drones and stuff and you know you see companies like Andrew and Palantir that explicitly recruit that way. You know they say come here if you want to build weapons and they're able to get a lot of very very talented engineers that way. You know they have they have a lot of um you know they they have a lot of strong uh uh uh talent pool there um I mean there are people who are willing to do that work you know in terms of Amazon you know Amazon then uh becomes a uh one of the primary contractors for the JWCC the joint warfighting cloud capability which is just sort of like a a dod-wide cloud computing um contract so you know they are Amazon is very much in the business of the US military you know uh and and you know as it pertains to Israel they're also um one of the main uh contractors on Project Nimbus which is the cloud computing uh platform that supports the whole Israeli military but including the military the Ministry of Defense and the IDF and the Air Force and so forth.

SPEAKER_01 46:45

Yeah and and there was a lot of controversy around that I believe last year when it was revealed uh that that a lot of data that had been collected on Palestinians was stored actually in Microsoft servers um in Europe, I believe in Netherlands uh in the end, right? And you know I know that Microsoft increased the security around its data centers in in response to that because it was worried about something happening which you know was certain certainly came to mind to me when I saw uh Iran actually target data centers, uh Amazon in this case of course.

SPEAKER_02 47:13

Yeah I mean uh so so the Amazon data centers that were targeted by Iran uh that that you're meant you know that we talked about a little earlier um again unclear if there was any military disruption there. Um I mean all that was publicly reported was you know disruptions to things like banking apps and you know food deliveries, you know, consumer um use cases. But Amazon and Google data centers within Israel are hosting Project Nimbus. They are hosting military workloads for the IDF and not just for the IDF but for um I've reported the uh uh Israel aerospace industries and Raphael which are the two main um Israeli state owned weapons uh manufacturers that have that have you know built the bombs that are then that have been used to destroy so much of Gaza I I I have to imagine that Iran hasn't uh would love to um strike those but obviously being within Israel it's much much harder to get through than um it is in you know Bahrain but you know those those would be certainly be squarely military targets.

SPEAKER_01 48:11

Yeah and and you mentioned before or maybe you didn't but you mentioned in your piece that uh you know there were there was a news agency in Iran that effectively listed a bunch of data centers dozens across the the Middle East as potential targets right uh many of them belonging to Amazon, Microsoft, these these US companies.

SPEAKER_02 48:28

Yeah um and you know the the the article that I wrote was sort of about the you know quote unquote legality uh under international humanitarian law or you know what's referred to as the laws of armed conflict. You know that that was something I was interested in is you know according to these laws and I want to use the word laws loosely here because um what we refer to as international humanitarian law is just like a patchwork of different treaties that countries have um either ratified or not. In the case of the US and Israel and Iran actually they have not ratified some of the most relevant um frameworks so you know war is basically lawless right like countries do what they can get away with. But according to the letter of the law a data center that carries uh military workloads on behalf of a uh you know a state military uh could be you know could be a a legitimate target for bombing or for a drone attack or or or or what have you um and so you know I think that uh particularly the the NIMBIS data centers in Israel uh I I think I imagine would have a huge target on the on their on their roofs.

SPEAKER_01 49:36

Yeah definitely there's one other piece of this that I wanted to get to with you before we start to wrap up our conversation and that is really you know you see these data centers in Bahrain and the UAE getting targeted by Iran and it's hard not to think about the fact that the Gulf has been trying to reposition itself for a number of years now as a major tech player and more recently a major AI player. Trump did this big uh you know tour through the Middle East last year and secured a bunch of commitments not just from US companies investing in data centers in Gulf countries like Saudi Arabia and and the UAE, but also a lot of investment then coming back to the United States for the tech center totally showed up in the mail. But yeah they have a lot of uh infrastructure to rebuild now as a result of a US war right um but I I wonder what you think that you know maybe this war but but even just these attacks on data centers in the region mean for for this ambition to become you know a major kind of tech region, a major tech player uh on behalf of the the Gulf monarchies?

SPEAKER_02 50:42

Well it's certainly shown how vulnerable they are and we and will be for the foreseeable uh future um I mean like I mentioned before the the the actual computer hardware in question here is very delicate and sensitive. I mean even a small uh drone carrying a relatively small payload if you set off the sprinklers in the data center right like uh your your your toast um uh a small fire could shut down a data center um on on on its own um you know an electrical outage right I mean these things are vulnerable they're sensitive they're delicate and I think that um you know well I think obviously the countries that we're talking about here want are sort of envisioning a kind of like post-oil economy right like how you know I mean this this sounds like a TED talk but like you know if data's the new oil right like maybe it makes sense to uh gradually uh you know uh uh transition from you know digging wells to building these data centers. I mean I can imagine that being in like someone's PowerPoint. It would also give you a lot of um you know it's a way of linking further linking your economy with the West you know with the US um and getting yourself in the good graces of uh uh the the the you know the White House and um you know US American business leaders but yeah I mean I think it reveals that this is not going to be an easy thing if there's uh a war. Um you know if if there is a anytime there's a war in the Middle East which is certainly not going to uh be a possibility that goes away anytime probably within our lifetimes um these are these I think have been revealed as very soft targets and that gives a lot of leverage to a combatant on either side right like um it's hard enough to protect military infrastructure you know like a uh to protect like a base right or uh oil tankers now you have to worry about these massive massive massive buildings that are just literally giant targets um I I think it it has you know that that was maybe not an aspect of this that was thought through entirely on the other hand this probably opens up a huge whole new economy of data center based counter drone weapon systems I mean there I have to imagine there will be a feeding frenzy around um you know uh uh buying countermeasures for that kind of thing but um it certainly complicated the plans of all those Gulf countries yeah yeah my god I I hadn't even considered the uh the new market that would create but I think you're absolutely right there there's gonna be uh I I mean I I'm I'm the pitches are probably happening right now all around the world you know oh I have no doubt like you know you even even you think uh it's not related to data centers but so quickly after the war happened and and after Iran started bombing uh you know Gulf countries and and infrastructure and and things there, you know, Ukraine was there right away to start offering its you know kind of anti-drone technology that it has developed in in recent years um against Russia, right? Yeah I mean I I securing something so large is I think I don't know I don't want to say impossible but very very very difficult. And anytime there's a very very difficult military engineering problem there's a lot of money to be made um at at least in in at least trying to solve it. So yeah I I mean I there are already firms that advertise sort of uh data center uh uh defense systems but again I I think that is going to be a uh uh a a a big there'll be further investment in in things like that I don't think it'll be sufficient though. I mean I think at the end of the day if you really want to um set one of these things on fire uh you will probably find a way.

SPEAKER_01 54:18

Yeah uh I I completely agree with you right you know they're it's pretty easy to to do something to one of these data centers uh to make it so that they are disrupted so that there's something happening.

SPEAKER_02 54:28

Maybe they'll start building them underground you know like like um like you know like missile bases are or or you know they'll be there'll be some attempt to sort of uh harden them from from airstrikes only adding to the colossal expense uh required to build one of them I mean they're already huge huge uh undertakings if you have to now guard against aerial attack um that makes it even you know even more of a pain in the ass. I mean something something I also referred to in my the the story I did about this was you know I I I um I have to imagine that this would be potentially concerning to people who live near data centers. You know it it's always very opaque as to which data centers are carrying which workloads but you know Amazon all all the big hyperscalers have DOD specific data centers uh that they operate that are you know exclusive to um to the Pentagon. I mean Amazon for instance has dedicated DOD data centers in uh Virginia and uh in the Midwest and uh you know I I I can imagine people feeling a little uneasy living next to again a colossal a like literally enormous bombing target. And I think as time goes on and and um you know these technologies get more and more deeply integrated into militaries, they only become more attractive targets.

SPEAKER_01 55:44

Yeah it's it it's like a separate issue than what I was thinking about when you know it was revealed that Microsoft was storing this data on Palestinians, right? In that case I was like, okay, so these Microsoft data centers might become targets for like you know protesters or or individuals that want to you know bomb them to try to stop them from aiding uh the genocide in Gaza, right? And I I hadn't really considered the flip side of it where you know it's not some individual doing it, but you know an actual state might be targeting these data centers because uh you know they're aiding in a military campaign against their their territory. That was not really on my on my radar is something to even expect.

SPEAKER_02 56:29

I mean in any like World War III scenario, Northern Virginia and DC are gone immediately, but I you know I I think that the concentration of data centers in northern Virginia probably makes it one of the most bombable places uh in in in you know in in the world I mean that is that is like the the beating heart of um the American military uh the the the computerized arm of the American military um and anyway and the the American economy um but uh yeah I have to I have to imagine that would be one of the first things to get hit.

SPEAKER_01 57:05

Yeah uh I I would agree with you and yeah it's making me think about this in a in a whole different way now um yeah it's a it's a fascinating subject unfortunately you know regarding an absolutely devastating consequence for so many people right whether it's we're looking at the people in Gaza or now the people in Iran um you know who who are on the other side of these missiles uh or or bombs or whatnot whose locations are being determined by uh these AI systems right and and they are feeling the brunt of this certainly we can complain about gas prices or or things like that but it's like you know the that's that's the real um consequence of what's happening.

SPEAKER_02 57:46

To be subjected to any of this is unim is an unimaginable you know cruelty and nightmare. I I I I literally cannot imagine it. But yes, I mean to to to I I think the real horror is that it is being done at a you know as as that Washington Post story you sort of hinted nodded um towards it's being done at a deliberately accelerated clip. I I don't think there's any way well I don't think there's any way to practice war in a humane manner but I don't think there I don't I think that a fast pace and a concern for civilians are mutually exclusive. I think caring, you know look looking out for um or or trying to take precautions around killing innocent people cannot be done when you're trying to go as fast as possible.

SPEAKER_01 58:32

Yeah especially when you're looking at some of the people who are prosecuting this war, right? They don't seem to be the type of people who are thinking very much about the collateral damage and the people on the other side of these bombs and no certainly not no to say the least yeah yeah no it's it's terrible. But I really appreciate you coming on the show to talk about all of this with us because I think it's a topic that we really need to be paying attention to and unfortunately it's one that we're probably just going to keep seeing evolving in the years to come. So yeah keep up the the fantastic work. Obviously I highly recommend people read your work certainly read this piece but but the other work that you're doing um and thanks so much for coming back on the show.

SPEAKER_02 59:11

Happy to join you and um happy to come back anytime. Thank you.

SPEAKER_01 59:17

Sam Biddle is a technology journalist at the Intercept tech won't save us is made in partnership with the Nation magazine and it's hosted by me, Paris Marks. Production is by Kyla Houston Tech Won's Save Us relies on the support of listeners like you to keep providing critical perspectives on the tech industry. And for our sixth anniversary you can join hundreds of other supporters by going to patreon.com slash tech won't save us and making a pledge of your own to help us hit our goal of a hundred new supporters this month. Thanks for listening and make sure to come back next week