Lizzie O'Leary (2:57)

Secretary Hegseth reportedly also told Anthropic to make a decision by 5pm Friday. Have you ever seen anything like this before?

Shira Frankel (3:07)

So the Pentagon over the decades has had, you know, disagreements with defense manufacturers and contractors and certainly other branches of the government have had disagreements with ambassadors, you know, and with other countries leadership. We've seen a lot of kind of emergency summits called. I don't think I've ever seen a technology company, let alone an American technology company called in in this kind of way and really pressured in this kind of way to do business with the government.

Lizzie O'Leary (3:37)

Anthropic is one of the few major AI companies that is, at least publicly and we'll put some wiggle room in there asking for specific guardrails around their tech and then the administration seems to vehemently against that. But Anthropic also has these well regarded models. So there's both this issue for the Pentagon where they are threatening Anthropic but also kind of need them. I wonder when you look at this situation, who you think has more power in this moment?

Shira Frankel (4:13)

That's such a good question. I think that we're in a situation where the power is somewhat evenly matched in that they need one another. And there's this really this deeply political fight happening over a technology that the rank and file of the Department of Defense are using. They love, they say it's really important to them and they're really hopeful they can continue to work with, you know, Anthropic wants to work with the US government. They've talked about this publicly. It's great for their portfolio as a company. They want this contract and the government wants Anthropic. The Department of Defense specifically wants Anthropic. Its AI models are working for them. And so we have a situation where instead of the two sides coming together and just saying we need each other, we're working together, this is great. We instead have, you know, these political ideals coming in and Hegseth really pressuring Anthropic.

Lizzie O'Leary (5:11)

Then just as we were putting this show to bed, Anthropic issued a statement saying the DoD's threats, quote, do not change our position. We cannot in good conscience accede to their request. Today on the show Anthropic vs. The Pentagon. Is this a cut and dried standoff about ethics or a more nuanced fight in the court of public opinion? I'm Lizzie o' Leary and you're listening to what next TBD A show about technology, power and how the future will be determined. Stick around.

Shira Frankel (5:49)

Foreign.

Lizzie O'Leary (5:55)

This episode is brought to you by Duck AI, a new product from DuckDuckGo. It's 2026 and the news is full of AI. It can be a little overwhelming, especially because it's easy to wonder if what you are asking a chatbot is actually secure. That's why DuckDuckGo built Duck AI. Duck AI is designed to let you chat privately with the same AIs you might already be using. Plus, it' completely free. No signups, no subscriptions. Duck AI promises to protect your info from hackers, scammers and data hungry companies. And it's from DuckDuckGo, the company that's been protecting your privacy, not collecting your Data for nearly two decades. If you want to use AI but you're concerned about privacy, visit Duck AI TBD today. That's Duck AI DBD from DuckDuckGo, where AI is always optional and designed to be private.

Farnoosh Tarabi (6:52)

Hi, this is Farnoosh Tarabi from so Money with Farnoosh Tarabi and today I want to talk to you about Boost Mobile Quick Money Tip Stop paying a carrier Tax if your phone bill feels trapped in a pricey plan, this is your sign to unlock savings. Boost Mobile helps you reset your spending with the $25 Unlimited Forever plan, you can bring your own phone, pay $25 and get unlimited wireless forever. And that simple switch can unlock up to $600 in saving. That's money you could put towards paying down debt, investing or something that actually brings you joy. Those savings are based on average annual single line payment of AT&T Verizon and T Mobile customers compared to 12 months on the Boost Mobile Unlimited plan as of January 2026. For full offer details, visit boostmobile.com Over

Advertisement Voice (7:42)

90 of the top 100 US accounting firms trust Bill to simplify and secure bill pay. That's proven financial infrastructure built on over a trillion dollars of secure payments. Visit bill.com proven for a special offer,

Lizzie O'Leary (8:02)

I think we should back up and set the scene a little bit for people who have not been following this as closely as you have. How would you describe Anthropic through the lens of the other AI companies that are out there right now?

Shira Frankel (8:17)

Anthropic of all the Silicon Valley AI companies had kind of set itself up as the more moral, more ethical, more more more really worried about security AI company. It was set up by a group of people who left other AI companies because they wanted to work somewhere that had more guardrails. Early on they were among the first to kind of raise their hands and say, hey, AI industry, let's all get together and come up with a set of principles that we can all agree to about what AI should and shouldn't do. Safe AI looks like they've also been pretty outspoken that they won't do business with China. You know, they've, they've really taken a stance on a lot of things that have set them apart from other AI companies in Silicon Valley.

Lizzie O'Leary (9:01)

They've also been making news for months. They, I guess you could say turbocharged. The practice of vibe coding, which is just like coding, having the AI do it, even if you don't really know what you're doing. With the release of CLAUDE code, they are consistently putting out new AI agents. Do you think their models are regarded better than those of OpenAI and xai?

Shira Frankel (9:24)

I wouldn't say. I think all these models, it really depends on who you ask. I've spoken to people who prefer Claude over OpenAI's ChatGPT. I've spoken to people who think that Google is ahead of them in some aspects of AI. These companies, it really is an arms race between them in terms of who can create the best AI model. And unless you are like extremely in the weeds and have personal preference, I think much of them for the average person will feel the same. Xai, I think being an outlier from the people I've spoken to who work with it say that it is not as good of a product as the other three.

Lizzie O'Leary (10:00)

How much work does anthropic do for the DoD and what exactly do they do for them?

Shira Frankel (10:07)

So Anthropics. And for this I can back up a little bit. Anthropic is really in this pilot program with the DoD that started almost a year ago now where the DoD said we want to incorporate AI, we want to bring it in, and we want to use it for everything from analyzing meetings to all this signals intelligence that the Department of Defense collects from other countries. It's incredibly labor intensive to go through the amount of data that the Pentagon has, and that's sort of exactly what an AI model is good at. It's good at quickly looking at data, analyzing it, drawing conclusions, creating summaries and showing patterns. And so they put out this call out and Anthropic was really the first to work with them and also the first to be integrated through the Pentagon systems, through their work with Palantir, which is a data analytics company. So Anthropic comes in and it quickly shows that it can be useful on all these Data systems created by Palantir.

Lizzie O'Leary (10:59)

Can you explain the main sort of sticking points for Anthropic in its negotiations with the Pentagon?

Shira Frankel (11:08)

The sticking points that we're seeing right now really only emerged because in January, January 9, Secretary Hegseth puts out this MEM and he says, we, the government, think AI is critical for national security. We want it across the military, we want it across the Pentagon. We are in a arms race against China and Russia for who can develop the best AI. And it's really a call to arms across the government to incorporate AI. And one part of this says, in order to beat our adversaries at AI, in order to be the best at it, we need all these tech companies, including Anthropic, to give us unfettered access to their AI. We want no restrictions. We, the Pentagon, want to decide how we use it, the best way to use it. Autonomous weapons, drones, the whole shebang, Right? Like we, the Pentagon, want to decide how our AI is going to be used in the military. And not just Anthropic, but all the tech companies see this memo and they kind of are like, oh, okay, well, they're getting really serious about AI. We need to renegotiate these pilot programs, these contracts with the government. And because Anthropic is the tip of the spear and working already on classified systems in the Pentagon, they're widely distributed through the Pentagon. They're kind of the first to stand up and say, well, we do have guardrails. We're not going to give you this unfettered access you seem to be asking for. We have two real sticking points that we want written into this contract. The first is that Anthropic says we don't want mass. Our technology used for mass surveillance of Americans. And the second is we don't want our technology used in what they call kinetic autonomous systems, which I'll decipher that a weapon with no human in the loop. So any type of missile, any type of, you know, drone that would fire or would take action without a human being being looped into that decision.

Lizzie O'Leary (12:57)

And that, to me, is, I think, something that really, even if the rest of this is confusing an everyday person can understand, I do not want to this drone to fly over here, analyze the data from its cameras and fire a missile without a human being saying, oh, that is a gathering of people who want to do bad things, or that is a wedding party.

Shira Frankel (13:27)

Right? And Anthropic, really, what they're saying is we don't think our technology is ready for that. We're not there yet. They're kind of putting a bit of the brakes on this. And the Pentagon's answer to this, to be clear, is cool. But you don't get to be the ones to decide this. We want to be the ones that decide when an autonomous kinetic weapon happens and when it doesn't, the Pentagon is saying, we don't want a private company being the one to, you know, check the box and say, okay, now the technology is ready.

Lizzie O'Leary (13:57)

Which, you know, in theory these things make so much sense.

Shira Frankel (13:59)

Right.

Lizzie O'Leary (13:59)

Like both of these arguments sound reasonable in practice. They are about something that is potentially lethal and potentially pretty game changing in the way conflicts are fought.

Shira Frankel (14:11)

Absolutely. And I, as I've been talking and thinking about this story over the last week and a half, the thing I keep telling people is like, really, this is a fight over the future of war and what war is going to look like. We are in a moment right now where every country, including China and Russia and every country with a modernized military is building autonomous weapons. They are trying as quickly as possible to figure out how their army can work with weapons that are AI backed, that don't require a human being to pull the trigger. And the US feels it must be competitive with this. It's really like another nuclear arms race. We just haven't called it that. And because you can't see the nuclear bomb being built, it's a little bit more esoteric. But this is where we're at, right? Everyone's building the robot army and now you have a company being like, hold on, we know we're in this arms race, but, but, but we don't know when we're going to be ready. And nobody is clear on when we're going to be ready. And we're worried about who's going to, who's going to be the one that says, yes, now we're ready for autonomous weapons.

Lizzie O'Leary (15:14)

So we have these two sort of competing ultimatums from the Pentagon. One is your supply chain risk. You can't do business with the federal government. The other is just kidding. You have to do business with the federal government and we invoke the Defense Production act and tell you what to do. If they did actually invoke the Defense Production act, what would that do to Anthropic.

Shira Frankel (15:37)

If they invoke the Defense Production act, it's an unprecedented situation. But what's happened in the past with that act is a company's been compelled to work with the government and they have been forced to provide their services, their technology, their products, whatever it is. To the government for national security reasons.

Lizzie O'Leary (15:55)

Why not just go to another AI company? I mean, the Pentagon has these pilot programs with other companies. They just came to an agreement to use Xai on classified system. Why not do that?

Shira Frankel (16:07)

They will try. The Pentagon has already signed XAI as a very willing partner. And Xai has said that they're kind of willing to come to whatever terms the Pentagon wants. They don't have the same guardrails, the same restrictions. They're, they're giving the Pentagon a carte blanche, unsurprisingly. This is a company led by Elon Musk, who has been very willing to have his technology used by the government in whatever way he sees fit. Right. So that's, that's done. But if you talk to rank and file people at the Pentagon, they're like, yeah, we don't, we don't really want Xai. We don't think the system is as good. The bit that we've played around with it, it doesn't work in the same way. The other partner that they're really heavily pursuing right now is Google. And from everything we've heard, Google is pretty close to signing a deal with, to work on classified systems and sort of do the work that Anthropic is doing. And Google wants this contract. Google has really aggressively pursued contracts with the government. It's just, let's say they sign tomorrow, right? Let's say the government makes good on its threat and they either boot Anthropic out or force them to work. And at the same time they announce, hey, we've also got Google. It's all going to take time. It's not like you are swapping in one thing for another with the ease of a push of a button. It takes time to learn a new system and to deploy it across the government. Time in which you're not analyzing the signals intelligence in the same way and doing all the other things that AI is good at doing. And so at a moment where the U.S. is, you know, threatening to go to war with Iran, lots of things are happening. Is this a moment where you really want to have a big transition?

Lizzie O'Leary (17:39)

Well, it sounds like that would give Anthropic some leverage here.

Shira Frankel (17:43)

It could. Or, you know, the US government could say, right, we're about to go to war with Iran. The Defense Production act is now in play and you have to work with us. And at the same time, we're onboarding Google.

Lizzie O'Leary (17:57)

When we come back, Wait, what do Americans think about autonomous killer robots?

Farnoosh Tarabi (18:11)

Hi, this is Farnoosh Tarabi from Sew money with Farnoosh Tarabi and today I want to talk to you about BoostBubble Quick Money Tip Stop paying a carrier Tax if your phone bill feels trapped in a pricey plan, this is your sign to unlock savings. Boost Mobile helps you reset your spending. With the $25 Unlimited Forever plan, you can bring your own phone, pay $25 and get unlimited wireless forever. And that simple switch can unlock up to $600 in savings a year. That's money you could put towards paying down debt, investing or something that actually brings you joy. Those savings are based on average annual single line payment of AT&T Verizon and T Mob, compared to 12 months on the Boost Mobile Unlimited plan as of January 2026. For full offer details, visit boostmobile.com hey,

Adam Grant (19:01)

this is Adam Grant, host of Ted's podcast Rethinking with Adam Grant. Let me share with you why smart finance leaders turn to Bill. They know that clarity isn't just helpful, it's strategic. As the intelligent finance platform, Bill uses AI to automate the busy work for nearly half a million businesses so they can focus on intentional growth, eliminate the friction and start scaling with the proven choice. Visit bill.compenven to talk with an expert about automating your business finances and get a $250 gift card as a thank you. That's bill.com proven terms and conditions apply. See Offer page for details.

Lizzie O'Leary (19:37)

You know, it's funny you mentioned Google because Google is a public company. Google has to answer to shareholders. If Google is going to agree to various DoD stipulations, that is something that its shareholders might not like and the markets might not like or like. But regardless, it's a different calculation for a company like Google than it is for Anthropic, certainly.

Shira Frankel (20:04)

And Google is a company that historically has had a workforce that has objected to some of its contracts with the government. Years ago there was a contract that Google was just pursuing they hadn't yet even signed. Yes, and you had project Maven and you had the Google workforce rise up against the company and say, absolutely not. We don't want to do this. And so I do think it'll be interesting to see whether Google's workforce once again does object to what, you know, they're signing. Or maybe they've changed. I mean, times have changed. More of Silicon Valley is willing to work with the government and to work in defense. And it's unclear. And then I will add, you also have OpenAI. OpenAI is another player in all this. They don't want to Be the first. As we can see, they're kind of lagging in back and kind of waiting to see what plays out with Google and with Anthropic. But they are another company that the government would be happy to partner with.

Lizzie O'Leary (20:53)

We've talked a bit about how Anthropic has positioned itself as the more ethical AI company that it says it wants regulation. Cynical question. Does Anthropic really believe any of this or do they win PR brownie points either way? Because they're out there saying that they do not want their systems used in this way. And if they get dragooned by the Defense Production Act. Well, they said it anyway.

Shira Frankel (21:24)

Yeah, you know, I'm not in their heads. I don't know whether they mean it or not. You know, their, their CEO has been really outspoken about the call for guardrails. And I think it's popular for every tech CEO right now to call for regulations. That's an easy one.

Lizzie O'Leary (21:40)

Which are, we should say, like, not gonna happen in the environment. So it's safe to call for regulation.

Shira Frankel (21:46)

Remind me, how long has Mark Zuckerberg been calling for regulations on social media? Eight years, nine years. Yeah, it doesn't happen and they all know it. I think that AI, you know, all these companies are really vying for talent. And one way to distinguish yourself and to draw the engineer you want to come work for you is to say, but look, we're the, we're the good guys. We're the, we're the AI company with safety guards and we're the ones talking about this publicly. And it could be another way to draw certain people to working at your company. And it could be real.

Lizzie O'Leary (22:16)

One of the things that I have heard from sources talking to about really more about the companies than this particular story is Anthropic has been beefing up its lobbying and policy shop and reaching out to a lot of former Democratic staffers. OpenAI has taken a somewhat different approach, has a slightly more Republican approach to this. There's this whole kind of behind the scenes perceptions push with these publicly facing Silicon Valley companies. And I wonder how you read this incident against the backdrop of who in Silicon Valley is vying for what power in the future.

Shira Frankel (23:03)

Yeah, I mean, these tech companies, their lobbying arms are vast and they've become more and more necessary as their technologies become more and more political. One thing I found interesting was people close to Secretary Hegseth were quoted as saying he didn't want Anthropic because it was quote, unquote, woke AI, which I'll just Be clear, that's a very nebulous term that nobody can define. I have yet to find a single person at the Department of Defense that can tell me what woke AI even means. But there's this sense of, oh, well, maybe anthropic is led by woke American, you know, progressives who live in California and who will code it to do something that is more. I don't know. I don't know what. I really struggle to come up with a single example of how this could play out. And yes, it's staffed by Democrats, and its lobbying arm has a lot of Democratic, you know, former Democratic Party operatives working on it, et cetera. And the politicization of technology that. Look, whether or not you want to see autonomous weapons is in my mind an important thing that should be discussed in war colleges and by people in the military and by people who can really strategically think about this. Whether or not the Pentagon should be able to analyze meeting notes and signals intelligence with AI seems pretty clear cut to me. I know every military that I, that I've spoken to, the British military, the Israeli military, they're all doing this. Yeah, analyzing a meeting note or analyzing a signal's intelligence with AI it shouldn't be a political question. It shouldn't require this idea of, like, is this technology company led by somebody who, you know, subscribes to the Democratic Party, the Republican Party?

Lizzie O'Leary (24:37)

Is there any sense of where the American public comes down on the idea of autonomous weapons?

Shira Frankel (24:45)

I don't think. I think that there's a lot of sci fi movies out there about autonomous weapons which, which explore the dangers of it. And I am, I'm a big sci fi reader. I read, I read a lot of sci fi. I really enjoy, um, and a lot of the sci fi books really explore this idea of. The problem isn't autonomous weapons in and of themselves. The problem is when you take a human out of the loop and it comes to the moment of a decision, can you trust a machine to make the right decision? Um, I think that idea is really interesting. And actually where. Where the people I've spoken to, where they fall on this, that I find most interesting is, is. Is that the problem isn't either one of those things. The problem is actually who pushes the button. And the minute you create the robot army, it might be that you create it for one administration, you create it for one prime minister in a specific country. But elections happen, revolutions happen, military coups happen. And once these weapons are in the hands of somebody who doesn't have to give orders to a human being and expect them to follow, but can give them to a machine which will follow. What does that say about the world we've created? What does that say about the future of warfare?

Lizzie O'Leary (25:51)

There's this paper that just came out from King's College, London, where three leading large language models, GPT5 2 Claude Sonnet 4 and Gemini 3 Flash, they all kept recommending nuclear weapons. This is like, oh, wait, guys, no, hold on. What's happening here?

Shira Frankel (26:11)

Yeah, it's. You know, I've covered war for years. I used to cover real life war in the Middle East. And the one thing that I always took out of war is that ultimately there is a hum, a moment, a very human moment where a person has to make a decision to pull the trigger. And you meet soldiers all the time that have been in those situations who talk about regrets they have or moments they feel proud of where they didn't pull a trigger. And they're so glad they did. Because the convoy was not in fact a group of terrorists moving ammunition. It was in fact a military, sorry, a wedding party moving from one place to another, a group of school children, whatever the case may be human beings, yes, they make mistakes, but also they have moments of moral judgment and moral clarity. And I think it's really concerning if the world is rushing into the situation where there isn't a human being facing another human being. And that's not the cost of war anymore.

Lizzie O'Leary (27:05)

That was one of the reasons I wanted to talk to you because you have this long history, not just of covering technology, but of covering war, of seeing what war does to people. How much have you gotten a sense? And I recognize that all the parties in this story are leaking, some truthfully, some to obscure their aims, what have you, that that aspect is being reckoned with by both the people at Anthropic and within the DoD.

Shira Frankel (27:41)

I have spoken to people at Anthropic that are thinking deeply about this and I've spoken to people at the DoD, especially like rank and file, like career DoD, people that are thinking deeply about this. I don't know if those people are in the same room thinking deeply together. I don't know that there has been a meeting of at least that I've seen or that the reporters have been able to get readouts on of these people getting together and being like, how do we move forward on this? What's the right way to really safely move forward on this?

Lizzie O'Leary (28:19)

Shira Frankel, thank you so much for talking with me.

Shira Frankel (28:23)

Thank you for having me.

Lizzie O'Leary (28:25)

On Friday afternoon, just before 4pm President Trump posted on Truth Social that he was directing, quote, every federal agency in the United States government to immediately cease all use of Anthropic's technology. Then he said there would be a six month phase out for the Pentagon and some other agencies which, and I'm just saying might be enough time for the various parties here to work something out. As for what all of this will mean in practice, we are going to be watching this story and we will keep you posted. Shira Frankel is a reporter for the New York Times. All right, that is it for our show today. What Next TBD is produced by Evan Campbell and Patrick Fort. Our show is edited by Elena Schwartz. Paige Osborne is the senior Supervising Producer for what Next and what Next tbd. Mia Lobel is the Executive Producer of Podcasts here at Slate. TBD is part of the larger what Next Family and we will be back with a special episode on Sunday. I'm Lizzie o'. Leary. Thanks for listening.

Advertisement Voice (29:40)

Blood Donation is Now More Inclusive More people are able to donate blood with the American Red Cross through FDA guidelines that eliminate eligibility questions based on sexual orientation. The Red Cross celebrates this historic change and welcomes those who may be newly elsewhere eligible to donate blood. There's a place for everyone in the mission of the Red Cross. The Red Cross is committed to achieving an inclusive blood donation process that treats all potential donors with equality and respect while maintaining the safety of the blood supply. Join us and help save lives. To learn more and make your appointment to donate blood, visit redcrossblood.org LGBTQ that's

Adam Grant (30:19)

redcrossblood.org LGBTQ hey, this is Adam Grant, host of Ted's podcast, Rethinking with Adam Grant. Have you heard of Bill? It's the intelligent finance platform that uses AI to help you avoid costly errors and optimize cash flow. In fact, Bill reports that over 90 of the top 100 US accounting firms trust them to manage, move and maximize money proven by over a trillion dollars in secure transactions. Eliminate the friction and start scaling with the proven choice. Visit bill.com proven to talk with an expert about automating your business finances and get a $250 gift card as a thank you. That's bill.com proven terms and conditions apply. See Offer page for details.