The Pentagon Wants an Obedient A.I. Soldier. Will It Get One?

Summary

The Political Scene | The New Yorker
Episode: The Pentagon Wants an Obedient A.I. Soldier. Will It Get One?
Date: March 18, 2026
Host: Tyler Foggatt
Guest: Gideon Lewis-Krause (Staff Writer, The New Yorker)

---

Episode Overview

In this episode, Tyler Foggatt explores the explosive conflict between Anthropic—a leading AI company dedicated to ethical and safe artificial intelligence—and the Department of Defense (DoD). The discussion, led by staff writer Gideon Lewis-Krause, examines recent allegations that Anthropic’s language model, Claude, was used in controversial U.S. military operations in Venezuela and Iran. The episode dissects how these events led to a breakdown in the relationship between Anthropic and the Pentagon, broader implications for AI governance, and the ripple effects throughout Silicon Valley.

---

Key Discussion Points & Insights

1. Initial Reactions: AI in Military Operations

• Increased Concern:
  • Gideon: “Oh, significantly more freaked out…” (01:50)
    Reporting on AI, especially its military use, amplifies fears: AI enables destructive capabilities on an unprecedented scale, fosters international arms races, and raises questions about control once these technologies are unleashed.

2. How Did Claude End Up in Pentagon Operations?

• Background:
  • Anthropic’s AI (Claude) was used by the DoD through Palantir—a company known for integrating massive, disparate data streams for military decision-making.
  • Gideon: “Claude was the first LLM that was certified for use on classified servers…” (05:20)
• Chain of Use:
  • The U.S. government accessed Claude via a Palantir contract, with Anthropic as a sub-vendor.
  • Users could select Claude through Palantir’s dropdown menu, and it was the favored model due to its performance. (06:20)

3. AI’s Role in the Venezuela and Iran Operations

• Venezuelan Operation:
  • Little concrete information available; initial awareness surfaced through rumor channels rather than formal reporting. (04:54)
• Iranian Operation:
  • Claude reportedly accelerated battle-planning and target selection, drastically reducing operation timelines.
  • Gideon: “That was able to generate, you know, a thousand targets in the first 24 hours.” (07:37)
• Risks of Automation and Mistakes:
  • Notable civilian casualties—like the assault on a girls’ elementary school—underscore the danger of AI processing faulty intelligence. (08:24)
  • Gideon: “If we had abstracted Claude out of this equation… Would a human have made the same mistake?” (09:01–09:24)
  • The key issue: would humans, given the same info, make the same errors as AI? AI might prevent some mistakes with more data, but it could also amplify them if the data itself is faulty.

4. The Anthropic–Pentagon Contract: Sticking Points and Breakdown

• Contract Terms:
  • Two core restrictions:
    1. Claude could not be used for fully autonomous weaponry (requiring “human in the loop”)
    2. Claude could not be used for domestic mass surveillance (an intentionally broad and undefined term)
  • Gideon: “…They did not want Claude used for fully autonomous weaponry, meaning without some human involved, and they did not want Claude used for domestic mass surveillance.” (15:40)
• Why the Split?
  • Emile Michael of DoD wanted standardized (and looser) agreements across all contractors, eschewing bespoke restrictions like Anthropic’s.
  • The Pentagon pushed for “all lawful uses” rather than observance of contractually stipulated red lines. (18:28–20:08)

5. Draconian Response: The ‘Supply Chain Risk’ Designation

• Pentagon’s Blacklist Move:
  • Rather than simply canceling the contract, Pete Hegseth (DoD) threatened to have Anthropic labeled a ‘supply chain risk.’
  • Gideon: "What Hegseth initially threatened... would have destroyed the company... [by] fully cut[ting] them off [from cloud and chip providers]." (21:14)
  • Ultimately, the threat was narrowed, but Anthropic countersued immediately, and is expected to prevail in court.

6. Normal Tech vs. ‘Weird’ AI: Fundamental Disagreement

• Two Worldviews:
  • DoD sees AI as “normal tech”—like a tank, gun, or plane—governed by existing laws and norms.
  • AI researchers and Anthropic see AI as fundamentally unpredictable, powerful, and opaque—even to its creators.
  • Gideon: “…if AI is like a normal technology, then it is like a tank, it is like a gun… The alternative view is… AI is so fundamentally different and transformative that in fact, actually, tomorrow is unlikely to resemble today.” (24:24)
• The Alignment Problem:
  • Root issue: Can we build something we can reliably control?
  • The Pentagon wants “a perfectly obedient soldier,” which experts say doesn’t exist.
  • Gideon: "Anthropic is saying back what you want doesn't exist and that you are opening a gigantic Pandora's box…" (26:25–28:56)

7. OpenAI Steps In—and Claims They’ll Solve It Technically

• Aftermath & OpenAI:
  • Following the Anthropic dispute, OpenAI (Sam Altman) swiftly signed a contract with the Pentagon.
  • Altman claimed to secure the same red lines via technical means (“safety stack”) rather than contract terms.
  • Gideon: "...we can offer a technical guarantee that it's not going to do certain things, which, like, flies in the face of what lots and lots of his own employees have said." (33:04–34:52)
• Skepticism:
  • The idea that AI’s behavior can be reliably controlled via technical means alone remains highly contested.

8. Industry Responses & Broader Implications

• Strategy Dilemma for AI Companies:
  • Should companies “go along” with government demands for some influence, or hold out and risk being replaced by less principled competitors?
  • Gideon: “By that logic you can kind of justify anything by saying that, like, the other guy is going to be worse.” (35:46)
• Tech Industry Pushback:
  • Anthropic has seen Silicon Valley rally in support, with amicus briefs and public solidarity.
  • Government threats may backfire by discouraging companies from working with the government at all—possibly leading to an innovation climate dominated by far less scrupulous players.
  • Gideon: "...if you're another tech executive, like, this has been a real warning shot, which is if you are doing business with the government ... the government will threaten to destroy you, which is, you know, really bad for the climate of investment..." (37:48)

9. Potential Outcomes and the Future

• Risk of a ‘Nightmare Technology Torment Nexus’:
  • If the government pushes out cautious companies, the field may be left to entrepreneurs eager to build whatever’s asked—regardless of the risks or ethical costs.
• Hopeful Signs:
  • Many companies are still aligning with Anthropic, resisting overly aggressive government intervention.
  • Gideon: "Lots and lots of tech companies have signed amicus briefs in support of Anthropic. So... there's something of a united front from Silicon Valley against this kind of unwanted interference." (40:14)
• Uncertainty Persists:
  • The lawsuit and regulatory battle between Anthropic and the DoD is ongoing.
  • The sector—and U.S. technological leadership—hangs in the balance.

---

Memorable Quotes & Moments

• Gideon Lewis-Krause on Accountability:
  “So often accountability is just like finding a scalp, right, where like, that's not meaningful accountability. In terms of what was the process that led to this catastrophic outcome that was just like we had an outcome and like that was the guy who happened to press the final button.” (11:44)

• On the "Alignment Problem":
  "The AI community is essentially saying, like, well, it's news to us that you guys suddenly solved the alignment problem. That like, you know how to control this thing because we build it and we don't know how to control it." (26:25)

• On 'Normal Tech' vs. AI:
  “If AI is like a normal technology, then it is like a tank, it is like a gun... [otherwise] tomorrow is unlikely to resemble today.” (24:24)

• On Red Lines and Safety:
  "We are going to make sure from a technical perspective that it does not cross these red lines… Now, it's anyone's guess whether Sam Altman really believes that..." (33:04–34:52)

• On Industry Fallout:
  "...if you're another tech executive, like, this has been a real warning shot, which is if you are doing business with the government ... the government will threaten to destroy you..." (37:48)

---

Key Timestamps

• 01:50 – Gideon describes heightened concerns about military AI.
• 04:54 – Chronicle of Claude’s use in Venezuela operation.
• 07:37 – Details on AI’s role in rapid military target planning in Iran.
• 08:24–09:24 – The accountability dilemma and ‘human vs. AI’ errors.
• 15:40 – Stipulations of Anthropic's contract with the Pentagon.
• 18:28–20:08 – Contract breakdown: push for uniform, less restrictive terms.
• 21:14 – Supplying supply chain risk details and company-destroying risk.
• 24:24 – "Normal tech" vs. “radically new” AI models.
• 26:25–28:56 – The alignment problem and AI obedience.
• 33:04–34:52 – OpenAI’s deal with the Pentagon and technical enforcement of red lines.
• 35:46 – Dilemma over whether ethical companies should capitulate.
• 37:48 – Broader tech industry and economic implications.
• 40:14 – Silicon Valley’s initial unified resistance.

---

Final Reflections

This episode offers an in-depth, clear-eyed look at the ethical and strategic fault lines emerging as AI becomes inseparable from national defense. It confronts big questions: Can advanced AI ever be a “normal” tool? What does real accountability look like when machines are decision-makers? And how will the standoff between tech and government shape not just U.S. policy—but the very character of American innovation?

Anthropic’s fight, supported by much of Silicon Valley, signals a critical juncture. The outcome could redefine the balance between technological progress, government power, and the safeguarding of human rights in the era of autonomous systems.

Transcript

[TurboTax Advertiser]

I don't know if you have this feeling, but I constantly find myself daydreaming about finding an expert who can solve a very specific problem in my life. And one specific problem that comes up every year is filing taxes. But luckily, finding someone who can solve that problem is easy. With TurboTax. They have dedicated tax experts who are there to help you every step of the way. With TurboTax Expert full service, you can match with an expert who will handle your taxes from start to finish, no matter how complex. You can even hand off all your tax stuff within the TurboTax app. And then you can see updates on your expert's progress while you go about your day. You can check exactly where your tax return stands right there on your phone. And then if someone ever says to you, man, I wish there were an expert who could do my taxes for me, you can say, oh, I know someone. Visit turbotax.com to match with your expert today. It's only available with TurboTax full service experts. The real time updates are only in the iOS mobile app.

[Tyler Foggatt]

Taking care of your eyes shouldn't be a hassle. That's why Warby Parker is a one stop shop for all your vision needs. Our prescription glasses and sunglasses are expertly crafted and unexpectedly affordable. Stop by a nearby store or use our app to virtually try on frames and get personalized recommendations. Did we mention we offer eye exams and take vision insurance too? For everything you need to see, head to your nearest Warby Parker store or visit warbyparker.com today. That's warbyparker.com. Hey Gideon.

[Gideon Lewis-Krause]

Hey Tyler.

[Tyler Foggatt]

Thanks for coming back so quickly.

[Gideon Lewis-Krause]

Thank you for having me.

[Tyler Foggatt]

Would you say that your reporting on AI has made you more freaked out about AI potentially being used in military operations? Or are you less freaked out?

[Gideon Lewis-Krause]

Oh, significantly more freaked out. Part of it is just that, yes, it seems to allow us to unleash the forces of destruction on a much greater scale. Then of course there's the question of like the game theory of this, which is like, if we are doing this, then certainly our adversaries are gonna be working on doing this kind of thing. And then finally there's the much more ticklish question of to what extent can we, like once we have trained it to produce this amount of devastation, to what extent are we really ultimately able to control it?

[Tyler Foggatt]

That's Gideon Lewis Krause, a staff writer at the New Yorker. Last month, Gideon came on the podcast to talk about Anthropic, a company that has committed to developing an AI that's grounded in the principles of ethics and safety. But just days after Gideon came on the show, reports began to emerge that the Department of Defense had allegedly used Claude Anthropic's series of large language models in military actions in Venezuela and later Iran, potentially without the company's knowledge. Recently, when Anthropic tried to put guardrails around the way that the Pentagon is using their technology, Pete Hagseth and other DoD officials bristled, leading to a wild escalation in which the Pentagon labeled Anthropic a supply chain risk, essentially blacklisting the company Anthropic is now suing. In response, I wanted to talk with Gideon about what we actually know about the role of AI in these recent military operations, how this conflict between Anthropic and the Pentagon took shape, and what it might mean for how companies handle similar pressure from the Trump administration. This is the political scene. I'm Tyler Foggatt and I'm a senior editor at the New Yorker. So let's rewind a little bit. So when you were on the show recently, I think like a month ago, you described Anthropic as this company that was essentially made up of lovable nerds who were deeply committed to ethical AI and then next thing you know, reports are coming out about Anthropic series of large language models, Claude being used in the Maduro operation in Venezuela and more recently in the military operations in Iran. Were you surprised when you started seeing these headlines, or did you always have a feeling that something like this was going to happen?

[Gideon Lewis-Krause]

I mean, something like this was bound to happen. There's no way it was not going to happen. And there's been a tendency, especially online to sort of say that Anthropic in particular was kind of hoist on its own petard, that it's like you guys have just been talking for years about how incredibly powerful this is. And so obviously, the government was going to cotton to this. And now the government has believed what you've been saying, and now the government wants it for itself.

[Tyler Foggatt]

So, just getting more into the nitty gritty, to the extent that we can, about how Claude has been used in military operations so far. I mean, what do we know, if anything, about how Claude was used in the raid that resulted in the abduction of Venezuelan President Nicolas Maduro back in January?

[Gideon Lewis-Krause]

I mean, we seem to know almost nothing. And the event that triggered all of this is that somebody, apparently, according to the government at Anthropic, called a friend at Palantir and said, hey, you know, we heard A rumor that Claude was somehow used in the Maduro raid. Do you have any idea what was going on? And then that got reported back into the White House.

[Tyler Foggatt]

And can you just remind us what Palantir is?

[Gideon Lewis-Krause]

Palantir is essentially an information synthesis company. And like, what they offer is an interface that allows for huge amounts of information from disparate sources to be fed into one place where one person can access whatever seems to be decision relevant at a given time. And the difference with CLAUDE is that CLAUDE was the first LLM that was certified for use on classified servers. So there was information that they could use CLAUDE to process that they could not use Claude's competitors to process.

[Tyler Foggatt]

I mean, I guess I'm wondering, like this idea of the Department of War using Claude, is that as formal? Because obviously there was this contract between the government and Anthropic, but is it the kind of thing where they're being encouraged to use this service? Or would it be more like if I just one day decided to like use Claude to try to edit a New Yorker piece and then like our boss found out?

[Gideon Lewis-Krause]

So the initial contract that Anthropic was involved in was through Palantir, that Palantir had signed a contract with or, or they've had ongoing contracts with the government. They had signed a contract in 2024 where anthropic was essentially like a sub vendor of that contract, which is like we have this system and this is a system for real time information analysis that's going to pull in, you know, climate data and satellite imagery and signals intelligence and all sorts of things and put it in one place so that one person, you know, using what's essentially just like workday software, can have an overview that helps them immediately synthesize like vast troves of real time information. And that that system allowed a user to just pick from a dropdown menu of different AI options. CLAUDE was one of them. And like, I had heard that Claude was kind of the overwhelming favorite for users of Palantir software because Claude was just the best at doing this stuff.

[Tyler Foggatt]

So, I mean, before we get deeper into the relationship between Anthropic and the Pentagon and how that fell so dramatically, I just want to talk a little bit more about how CLAUDE may have been used in military operations in Iran.

[Gideon Lewis-Krause]

I mean, we certainly, we've only had a little bit of reporting about how exactly CLAUDE was used in Iran, but it seems like the initial reporting that we've seen has suggested that it was used, for example, for target selection and that there was a quote I think in the Washington Post saying that the kind of battle planning that might have taken people, you know, months to put together was able to be done using AI as part of a broader system. It's not just like they're sitting down with a chatbot, like, asking questions like, where should we bomb that? It's incorporated into the system that Palantir uses. That's kind of a total information awareness system that. That was able to generate, you know, a thousand targets in the first 24 hours.

[Tyler Foggatt]

Of course, when, you know, we think of targets and faulty targets, like, the first thing that comes to mind is just the horrible assault on that girl's elementary school, which killed around 170 people. The reporting doesn't show that Claude was necessarily used to pick that target, but the reporting has suggested that that target was based on bad intelligence. You know, like, apparently, I think the school used to be a military base or was close to one or something like that. So it does kind of introduce this idea of like, AI using faulty intelligence to suggest a target, which seems like one of the kind of the big things that we would have to be worried about here. Right.

[Gideon Lewis-Krause]

Well, so to me, kind of the key question in all of this stuff that sometimes doesn't get asked when people really are quick to blame a technological solution for something is like, what is the relevant counterfactual here is that, like, if this was in fact based on outdated military intelligence, would. If we had abstracted Claude out of this equation.

[Tyler Foggatt]

Yeah. Would a human have made the same?

[Gideon Lewis-Krause]

Would a human have made the mistake? Yeah, we don't know. And would you know. So to what extent can you say that this was specifically an AI related failure mode? Because you could also certainly imagine cases in which a human would have acted on bad information and an AI which would have, in theory, had access to orders of magnitude more information, would have said, like, oh, no, hold, you know, hold up here. Like, that's not right. And so, you know, people who do AI safety and, you know, this has come up a lot in terms of keeping a human in the loop and what they call the kill chain. They love to point to this example of Lt. Petrov during the Cold War, who noticed, you know, potential launch of atomic missiles and decided that this had to be a drill and didn't initiate a counterattack and like, you know, potentially save the world from nuclear destruction. And, you know, this isn't, you know, often used as an act of like, like the heroic individual conscience that has, like, prevented us from destroying each other. And it certainly was. But, you Know, it's very possible to imagine a system in which, in which an AI, in that case said, whoa, whoa, whoa, chill out, this is clearly like just a drill, like this is not real. And that it would have made the same decision. I mean, I don't think based on the information that we have now, although this very quickly gets into like kind of the heart of the matter, whether we even know enough to say that there would have been an appreciable difference between, like how a human would have acted in some of these high stakes situations and how an AI would act.

[Tyler Foggatt]

I mean, the appreciable difference is accountability, right? And I guess, like this is kind of jumping to the next part of the conversation, which is about what sort of destroyed the relationship between Anthropic and the Department of Defense or the Department of War, whatever you want to call it. But just like it's maybe human and AI would have made the same mistake, but at least when it's a human who makes that mistake, it's like it's someone to point to. It's their fault. And it seems like the only reason why the government would be so interested in fully autonomous weaponry, which was one of the sticking points in the deal between Anthropic and the Pentagon, is that they are trying to, you know, sort of shunt that accountability. I mean, it seems like they kind of want like an AI scapegoat, right? Or is it just so much faster if, you know, Claude can just pick all the targets and execute on them?

[Gideon Lewis-Krause]

Well, in a sense there is an open question whether it's better or worse to have an AI scapegoat. I mean, so people talk a lot about accountability, which clearly is the crucial question here. But so often accountability is just like finding a scalp, right, where like, that's not meaningful accountability. In terms of what was the process that led to this catastrophic outcome that was just like we had an outcome and like that was the guy who happened to press the final button. So like, well, let's put his head on a stake. And in theory that represents accountability. In practice, like, is that the kind of, of accountability that we want? I'm not so sure.

[Tyler Foggatt]

So let's take a break and then when we come back, I want to talk more about the sort of war between Anthropic and the Department of Defense. This is the political scene from the New Yorker.

[Bloomberg News Announcer]

Get informed no matter where you are or what you're doing.

[Gideon Lewis-Krause]

Subscribe to Bloomberg News now and get news when you want it on your schedule. Big tech dominating our headlines. Another win for Donald Trump.

[TurboTax Advertiser]

A key inflation report these are short

[Gideon Lewis-Krause]

audio reports, five minutes or less that

[Bloomberg News Announcer]

bring you the latest headlines with context

[Tyler Foggatt]

24 hours a day.

[Bloomberg News Announcer]

Stay on top of the latest news

[Gideon Lewis-Krause]

from around the world. Get it on your smartphone.

[Bloomberg News Announcer]

Subscribe to Bloomberg News now on Apple, Podcasts, Spotify, or anywhere else you listen. This show is supported by Odoo. When you buy business software from lots of vendors, the costs add up and it gets complicated and confusing. Odoo solves this. It's a single company that sells a suite of enterprise apps that handles everything from accounting to inventory to sales. Odoo is all connected on a single platform in a simple and affordable way. You can save money without missing out on the features you need. Check out Odoo at o d o o.com that's o d o o.com the

[Quince Advertiser]

seasons are changing and it's time to take a look at your wardrobe. Personally, I want wearable, versatile pieces that last. And that's why I'm relying on Quince. Quince has the everyday essentials I love with quality that lasts. Quince works directly with top factories and cuts out the middlemen. So you're not paying for brand markup or fancy retail stores, just quality clothing. The cashmere is 100% Mongolian. It's the same stuff luxury brands use. The Pima cotton is long staple, which means it stays soft and doesn't pill. The European jersey linen is breathable and lightweight. Everything is built to hold up to regular wear and still look good. Stop overcomplicating your wardrobe. You don't need a closet full of options that eh kind of don't work. You need a few great pieces that consistently look good day after day. I don't know about you, but I want quality right down to my underwear and Quint has me covered there too. Right now. Go to quince.com politicalscene for free shipping and 365 day returns. That's a full year to build your wardrobe and love it. And you will now available in Canada too. Don't keep settling for clothes that don't last. Go to Q-U-I-N-C-E.com politicalscene for free shipping and 365 day returns. Quince.com politicalscene.

[Tyler Foggatt]

So in your recent piece you give a really thorough overview of this escalating conflict between Anthropology, Anthropic and the Department of Defense. As you mentioned earlier, it seems like the Department of Defense first started using Claude through Palantir and then I guess it was in July of last year, that Anthropic signed its own, was it $200 million contract with the Department of Defense. So what were kind of like the stipulations of that contract? Like, what did it look like and what were like, the parameters that that Anthropic initially set out for how the tech could be used by the Pentagon?

[Gideon Lewis-Krause]

So, I mean, we certainly don't know all the details of that, but we definitely know what the two ultimate sticking points were, which was that in this contract, there were two stipulations on Anthropic's part, which was they did not want Claude used for fully autonomous weaponry, meaning without some human involved, and they did not want Claude used for domestic mass surveillance. Now, the tricky thing is that domestic mass surveillance is not a legal term. It has no legal meaning. Basically, their worry was that there are plenty of legal loopholes in our privacy regime that people have been pointing out for a really long time where essentially there are certain things the government is prevented from doing, not because it's illegal, but because it's simply too cost inefficient to do. And if the government wants to assemble a profile of you, Tyler, they could assign a person to go through all of that data, find which data points correspond to you.

[Tyler Foggatt]

All the Taylor Swift videos.

[Gideon Lewis-Krause]

All the Taylor Swift videos, yeah. You unfortunately, have left a very long data trail, and they could compile a dossier about you if they wanted to, but they can't do it for 350 million Americans unless they hired millions and millions of people to be doing this all the time. Whereas with Claude, it would be pretty easy for Claude to just do that for everyone based on publicly available information. And they did not want the government to do that. That.

[Tyler Foggatt]

And I'm going to quote your own piece back to you, but I found this line to be, like, really illuminating in terms of, like, what it would actually look like. But you write a panopticlad could make tailored watch lists all day long, say, matching concealed carry permits with unpatriotic tweets, or cross referencing protest attendance with voter rolls, which, like, really helps you get a sense of, like, how bad and overwhelming it could be if Claude were actually used in this way. Okay, so they said no to mass domestic surveillance, and then they said no to fully autonomous weaponry, which is what we talked about earlier. And I mean, this was back. They signed this contract in July of 2025. So why is it that, you know, like, several months later, that contract is falling apart and they're all of a sudden fighting over these Terms that it seems like the Pentagon must have agreed to back then.

[Gideon Lewis-Krause]

So some of this is kind of murky and probably depends on whom you ask. But it seems like what happened is that Emil Michael, who is a former Uber executive, he was the one who was behind, like, using Uber to track journalists who were, you know, writing negative things about the company.

[Tyler Foggatt]

His reputation precedes him.

[Gideon Lewis-Krause]

He was brought in as an undersecretary for research and development, I think, at

[Tyler Foggatt]

the DOD, at the DOD.

[Gideon Lewis-Krause]

And, you know, according to the government side of the story, he basically looks at their existing contracts and he decides that he wants to render them uniform, that he doesn't want certain companies to have certain stipulations, that it's too hard for the government to be negotiating on an individual basis with each defense contractor, and that he kind of wants uniform contracts across the board. Now, from the alternate perspective on this is that, well, actually, plenty of companies already have lots of stipulations in their contracts. Like they are contracts. Contracts are things that are freely entered into and that you offer your terms and they, you know, the counterparty accepts them or doesn't accept them. And that, you know, it's not like when you buy an F16 from Lockheed Martin, you can do whatever you want with it. You have to still agree to, like, certain contractual stipulations there. But what Michael didn't like is that kind of according to White House sources or according to administration sources, he just wanted a clean thing where all defense contractors were going to have the same agreements, which is that the things that they sold to the government could be used for all lawful uses. Meaning we're the government, we follow the law. You know, we follow the law. Trust us, we always follow the law. And all we want is to be able to use the thing that you're giving us within the bounds of the law. The laws are, you know, a result of a democratic, democratically legitimate process. And we can't just be letting, you know, private individuals in San Francisco effectively setting their own laws based on the technologies that they're selling to us.

[Tyler Foggatt]

Totally. But it wasn't that the government was just going to rescind Anthropic's contract. It's that they had this whole other plan which was to label them a supply chain risk.

[Gideon Lewis-Krause]

Right? Yeah. So to take a step back, I mean, the cleanest possible thing, as even many conservatives pointed out, is that the government could have just walked away from the contract. They could have said, look, you know, we liked using Claude, it was useful, but we're not going to use Claude if you're going to be so meddlesome and interventionist here. And so like, let's just both walk away from this. And Anthropic would have lost $200 million, which is a rounding error on annualized, you know, their current.

[Tyler Foggatt]

What's their valuation?

[Gideon Lewis-Krause]

Well, the valuation's about $380 billion at last report, but their revenue like doubled in the last month. They're currently on path for an annualized revenue of like $20 billion. So $200 million, I mean, they make in a couple of days. They didn't care about the money. They weren't doing it for the money. So that certainly would have been the easiest thing to do.

[Tyler Foggatt]

But instead, what exactly happened? I mean, what led to Hegseth labeling Anthropic a supply chain risk? And what does, what does that mean?

[Gideon Lewis-Krause]

Well, that's a really good question. Now, typically, this is a restriction that has only ever been applied to companies, typically infrastructure companies that have ties to our adversaries. So, for example, Huawei, like we know that Huawei has a close relationship to the ccp. Huawei is a telecoms infrastructure company. It makes sense for the government to say, we don't want anyone we're working with to be using, you know, Huawei infrastructure because we don't know what it's going to be sending, exfiltrating and sending back to Beijing. So, so on one level, there would have been, okay, any government contractor, while in the course of doing their government business, cannot simultaneously be using an Anthropic product like Claude, meaning that, like, for your government workflows, you can't rely on Claude. And that would have been extreme and unprecedented, but not company destroying. But what Hegseth initially threatened was, you know, what other people have called, like a secondary sanctions regime or a secondary boycott regime, which would've interpret the law to mean we can prohibit any company that does any business with the government from doing any business with Anthropic. And this like, irrespective of what business they're doing. And this would have destroyed the company because, you know, Anthropic relies on these companies like Google and Amazon for cloud services and they rely on Nvidia for chips. And like, this would have fully cut them off and placed like, you know, what were clearly extremely punitive sanctions on the company. So that would have been existential. Now, people pointed out, you know, national security lawyers pointed out right away that this almost certainly vastly exceeded Hegseth's statutory authority and that that was never going to fly. And that finally, a week later, when Hegseth did make this official announcement that it was a supply chain risk, it was the narrower version of the law. And even that seems, you know, Anthropic filed a countersuit or a lawsuit two minutes later. And even that seems probably beyond what his authority was. And Anthropic will probably prevail in court.

[Tyler Foggatt]

I want to go back to an idea that you mentioned earlier, which is that the Pentagon is used to buying jets from Lockheed Martin or, you know, semiconductor chips. And usually it's a situation where, you know, the thing is just kind of sold to the government and the government gets to decide what to do with it. Basically, you know, it's this category that you refer to in the piece as normal tech. But obviously, or maybe not obviously, it kind of seems like AI is not normal tech. It's this thing that you wrote about, you know, so wonderfully, you know, about a month ago, just that it's this thing that's almost like weirdly sentient. And even Anthropic itself doesn't quite understand what exactly is happening with Claude and what its capabilities are. So I guess I'm wondering if you can just like walk us through this, you know, this normal tech idea and like, why it is that AI should maybe be considered something different and whether that is something that the government is ever going to be able to fully kind of like wrap their heads around.

[Gideon Lewis-Krause]

So the argument for AI is a normal technology, it's not saying that AI isn't potentially hugely important and transformative. It is saying that AI is potentially hugely transformative, but also not categorically different than anything else that is come before. That there is historical continuity here, that there have been other things like electricity and the semiconductor and the Internet and the private, you know, the home computer, all of which were transformative in their own ways, but none of which thoroughly transformed life and geopolitics as we know it. And you know what one government official said to me, like, what this fundamentally comes down to is a division over this question of normal technology, where if AI is like a normal technology, then it is like a tank, it is like a gun. It is something that behaves the way that you expect it to behave. It is a tool that is used in relatively predictable ways. It's not really going to surprise you. And so current laws apply and we can't have private individuals saying, like, no, no, no, we need, like, further restrictions. And so the alternative view is no, no, no. AI is something that you cannot place in a reference class with all of those other things that AI is so fundamentally different and transformative that in fact, actually tomorrow is unlikely to resemble today.

[Tyler Foggatt]

Is it possible that right now, as we speak, AI is not a normal technology, but eventually it could be one? I guess I just wonder, like you mentioned, like electricity and it's like, what if when electricity was invented, it was just like we didn't quite understand how it worked. Every time we tried to do it, there was like a 40% chance that you would electrocute yourself and die. I guess I wonder if like there are just like a lot of kinks that are still being ironed out, but that there will one day be a point at which we can regulate it or sort of treat it the way that we would a gun or a tank or a phone.

[Gideon Lewis-Krause]

So this is exactly, this is the crucial question. And in AI safety communities this is called the alignment problem. And kind of the classical way to look at the alignment problem is we want to make sure that an AI that is potentially one day as intelligent or more intelligent than we are remains aligned with, quote, human values. Now, as anybody could point out, what the hell are human values? That's not a load bearing concept. Different communities have different values, different people have different values. Now, another way to set aside the jargon of the alignment problem and this placeholder usage of a phrase like human values, it really is just a question of are we building something that we can reliably control or not. The government is essentially saying that's not a problem that we, like, we don't believe it's a problem, it's a normal technology. Like one administration official compared it to a computer virus and he was like, we used to have problems with computer viruses. And then we figured that out and now we don't really have problems with computer viruses anymore. And that all of these issues of control are actually just like, like it's going to be a long game of engineering whack a mole where when we see behaviors that we don't like, we're going to patch them and we're going to be fine. And the AI community is essentially saying, like, well, it's news to us that you guys suddenly solved the alignment problem. That like, you know how to control this thing because we build it and we don't know how to control it. And it basically seems like the government is saying we want an AI that is going to be a perfectly obedient soldier, an AI that is going to follow all orders and carry them out in the way we would it to carry out those orders. And Anthropic is saying back what you want doesn't exist and that you are opening a gigantic Pandora's box by saying you essentially want an AI with all the safeguards removed because you basically just trust that it's going to do what you want. And the analogy, which I don't think is at all far fetched, is that this is an administration whose chief and perhaps only principle is loyalty. And they're essentially saying, look, we have an administration full of loyalists who do the things we tell them to do, and if they don't do the things we tell them to do, we get rid of them and the AI is going to be treated the same way. We are going to get an AI that is perfectly loyal and we want you to give it to us. And Anthropic keeps being like, it just doesn't work that way.

[Tyler Foggatt]

In a minute I want to talk about what this means for the rest of the tech industry and for other AI companies, including OpenAI, which very soon after the deal between anthropic and the DoD fell apart, OpenAI signed its own contract with the Pentagon. This is the political scene from the New Yorker. Wired has always put a microscope on the people, power and forces shaping our world. Uncanny Valley brings that same fearless reporting straight to your feed. Is Doge finally over? Will AI actually democratize American healthcare? Each week, Wired journalists from across the newsroom are going to unpack where politics, technology and Silicon Valley collide. From conversations with tech leaders across Silicon Valley, Internet fandom investigations, and government crackdowns on rigged gambling, we're taking you all over the news cycle, going straight inside the priorities, pressures and power plays driving today's biggest decisions. Uncanny Valley tackles the questions keeping you up at night and helps make sense of the future taking shots shape right now. Listen to new episodes every Thursday, wherever you get your podcasts.

[Jason Adam Katzenstein]

Hi, I'm New Yorker cartoonist Jason Adam Katzenstein. I'm a short man with a small business, and that means I spend a lot of time hustling and trying to figure things out on my own. But now I don't need to spend my evenings guessing at tax forms or tracking down onboarding documents. Gusto handles all of that for me so I can spend time doing the thing I actually love, which is cartooning. Gusto is an online payroll and benefits software built for small businesses. It's all in one remote, friendly and incredibly easy to use so you can pay, hire onboard and support your team from anywhere. You've got automatic payroll tax filing, simple direct deposits, health Benefits, workers, comp 401k, you name it. Gusto makes it simple and has options for nearly every budget. One of my favorite subjects to cartoon is Sisyphus. Endlessly pushing that boulder up a hill and I felt like Sisyphus in the past with paperwork, forms and logistics. But now, thanks to Gusto, I don't need to live that existential dread. I can just draw about it. Try gusto [e-mail] newyorker and get three months free when you run your first payroll. That's three months of free [e-mail] new yorker one more time gusto.com new yorker

[Bloomberg News Announcer]

this show is supported by Blueland. Living a cleaner and more sustainable lifestyle by reducing single use plastics is a small action that makes a big difference and Blueland is on a mission to make it easy for everyone to make those sustainable choices. They believe that hard working clean products can be the norm, not the exception, so that you can do better for your family and the planet at the same time. From cleaning sprays and toilet bowl cleaner to dishwasher and laundry detergent tablets, Blueland's 100% microplastic free EPA safer choice certified formulas are safe to use around people, pets and plants. You'll love not having to choose between the safe option and what actually gets your house clean. Blueland products are independently tested to perform alongside major brands and the formulas are free from dyes, parabens and harsh chemicals. Plus Blueland is a certified B Corp and certified cruelty free by Leaping Bunny. Their formulas are EPA Safer Choice certified and many have received the Gold Material Health Certificate from Cradle to Cradle. And right now Blueland has a special offer for listeners. Get 15% off your first order by going to blueland.comprx you won't want to miss this. Blueland.comprx for 15% off. That's blueland.comprx to get 15% off.

[Tyler Foggatt]

So given everything that you just said, I guess like how worried should we be that after the contract between the DoD and Anthropic blew up, OpenAI signed one right after?

[Gideon Lewis-Krause]

Well, I mean according to Sam Altman himself, he was already negotiating with Emile Michael on the Wednesday before the the deadline. So we can basically infer that even if he claims he was trying to de escalate the situation, that he also was giving the DOD a way out of continuing to negotiate with Anthropic. And he basically said we got the same red lines that Anthropic wanted. That also we got the Pentagon to agree to no autonomous weapons and no surveillance but we did it in a different way that they were doing it by contract, whereas we're not going to do it by contract. We're going to have the contract that the government wants, which is this all lawful uses contract. But the way we're going to handle our red lines is we're going to bake it into ChatGPT itself. So it's going to be part of what he called our safety stack, which means the government can use it however it wants, but we are going to make sure from a technical perspective that it does not cross these red lines, that it does not compile dossiers on every American or function as a part of a fully autonomous kill chain. Now, that is effectively him saying the same thing, which is basically like we solve the alignment problem, that we are going to be able to guarantee that this performs the way that we say it's going to perform. Now, it's anyone's guess whether Sam Altman really believes that this or not. Because he started OpenAI because he didn't necessarily trust that AI was being developed safely and responsibly, and now he's basically saying, like, we can offer a technical guarantee that it's not going to do certain things, which, like, flies in the face of what lots and lots of his own employees have said.

[Tyler Foggatt]

So there's OpenAI. And then in your piece you also mentioned that there were these rumors that Altman's federal contract, which he'd never actually seemed to want, was just keeping the seat warm until GROK dropped the Hitler Causeway play in favor of functional competence. And I guess I'm just wondering, like, given that there are these other AI companies that are clearly willing to do business with the government, and given that the whole reason why Anthropic got involved with the Pentagon in the first place was that they sort of understood that it was, it was inevitable that the government was going to be interested in this technology. And at least if they were the ones to provide it, they could maybe try to do it in a safer, more ethical way, I guess. Do you think that there's an argument that they should have just kind of gone along with it to have some control over the situation as opposed to letting the government just do kind of whatever it wants with Altman and perhaps later with grok, which seems like maybe the worst case scenario?

[Gideon Lewis-Krause]

Well, okay, a few things. One is that by that logic you can kind of justify anything by saying that, like, the other guy is going to be worse. And in this particular case, if the government's end game is to ultimately, like, give this contract to Elon Musk and Xai, which it may very well be. I mean, plenty of inside sources at the Pentagon have gone to newspapers to say, like, we're not using grok, is not ready because not good, and like, we don't want to use it because it doesn't work well. And essentially, you know, Elon Musk has said that it'll be ready for use on classified. I think it was maybe already approved for classified systems, but that it will be fully ready for deployment by the end of the year. But the thing is, he's also lost, like, almost all of his co founders and engineers. So it's not really clear exactly how they're going to make GROK good enough to be useful when they just don't have the engineering talent to be doing it. So, you know, one calculation Anthropic could have been making is, well, yes, in theory, GROK would be a much worse replacement for us, but in practice that's just never going to happen because GROK is never going to be good enough for the government to use.

[Tyler Foggatt]

On that note, I guess, like, maybe this is like a good place to conclude, but what is your sense? And this is just, just really asking you to speculate. But, like, how do you think that this whole episode is going to affect how corporations and industries work with the White House more generally? At first I was like, oh, I wonder if like a bunch of companies are just going to fall in line whenever the administration starts to flex their power. But then you also see how, you know, people are leaving OpenAI. They've been getting a lot of bad press. Anthropic seems to kind of be having like a, you know, like a real moment right now. Like, a lot of people are incredibly like, a lot of people, like, subscribe to Claude, you know, right after news of this came out. And so I guess I wonder, like, if anything, if it might have just like a cooling effect on companies wanting to enter into government contracts or just increase the number of companies that are willing to kind of stand up against the government because they see that there can actually be benefits from that.

[Gideon Lewis-Krause]

Well, I mean, people have certainly drawn parallels to the law firm situation when Paul Weiss bent a knee when Trump insisted. But Wilmer Hale or whatever it's called now didn't. And actually Wilmer Hale are the attorneys for Anthropic in this case. So there's one question about, you know, there's kind of a collective action question about, like, if one company stands up, can they like, have some Kind of Spartacus moment where everybody joins them. I mean, the other variable here is inside the administration that there are big, big, you know, the kind of maga, like new tech. Right. Element of the administration. The last thing that they want to do is like, completely cripple the American economy and the tech economy by discouraging these companies from participating in government contracts. But if you're another tech executive, like, this has been a real warning shot, which is if you are doing business with the government and there is anything that you are refusing to do, the government will threaten to destroy you, which is, you know, really bad for the climate of investment and is discouraging people from participating, you know, in government contracts, specifically defense contracts. So the government is in a very difficult position here where, you know, they could end up just remaking the tech sector in their own image based only around people like Palmer Luckey and people who will like, happily build whatever nightmare technology torment nexus the government wants them to build.

[Tyler Foggatt]

CEO of Andura. Kind of like the big. The like crazy war drones, right? Yeah.

[Gideon Lewis-Krause]

But you would have to think that, like, some of these people in the administration, like, understand that that is going to have, like, very perverse effects on the kinds of technology that is available to you and that there are a lot of potentially very bad outcomes for America and for American technology if you. You're ending up in essentially like a Chinese or Soviet command economy. That it.

[Tyler Foggatt]

But for now, at least, it seems like anthropic. Like, because there was this real concern when it seemed like Hagseth was just going to be able to blacklist them kind of like entirely, that like, you know, the US has lost the tech cold war with China. Like, this is just a real destruction of, like, American innovation. It seems like they're going to be okay both through their lawsuit potentially, and then also, just like financially, they're able to kind of weather the storm.

[Gideon Lewis-Krause]

Certainly seems like it. I mean, this isn't over. But also like, one very heartening sign is that lots and lots of tech companies have signed Acts Lamachus briefs in support of Anthropic. So at least right now it seems like there's something of a united front from Silicon Valley against this kind of unwanted interference.

[Tyler Foggatt]

Thank you so much for being here, Gideon.

[Gideon Lewis-Krause]

Thank you, Tyler. It's always fun.

[Tyler Foggatt]

Gideon Lewis Krauss is a staff writer for the New Yorker. You can find his latest piece on the war between Anthropic and the [e-mail] this has been the political scene from the New Yorker. I'm Tyler Foggatt. This episode was produced by John Lamay, with mixing by Mike Kutchman and editing by Rhiannon Corby, as well as engineering by Pran Bandy. Our executive producer is Stephen Valentino. Our theme music is by Alison Layton Brown. Thanks so much for listening. We'll be back next Wednesday.

[Gideon Lewis-Krause]

From prx.