Anthropic’s Pentagon Problems

Summary

The Journal. — "Anthropic’s Pentagon Problems"

Podcast: The Journal
Hosts: Jessica Mendoza & Ryan Knutson
Date: February 23, 2026
Episode Theme:
A deep dive into the escalating conflict between Anthropic, a leading AI company, and the US Department of Defense (Pentagon), highlighting broader questions about the role of artificial intelligence in national security, ethics, and government policy. The episode explores the stakes, the culture clash, and the consequences of AI’s rapid integration into military operations.

---

Episode Overview

The episode examines how Anthropic, a company born out of concerns for responsible AI usage, found itself at odds with the Pentagon after landing a major military contract. Tensions erupted over the uses and ethical limits of AI technologies—especially following a high-profile US military operation that reportedly leveraged Anthropic’s AI. The hosts and Wall Street Journal reporter Amrit Ramkumar break down the unfolding standoff, its political undertones, and what it portends for the future of government AI adoption.

---

Key Discussion Points & Insights

1. The Defense Tech Summit & AI’s Military Integration

• [00:05–00:36] The episode kicks off describing a high-profile Defense Technology Summit in West Palm Beach, Florida—a meeting including Pentagon leaders and top tech figures, with the goal to modernize the military with advanced tech, particularly AI.
• Amrit Ramkumar: “The motivation is to embed the most advanced technology throughout the US Military.” (00:13)

2. Anthropic: Origins & Safety Focus

• [04:02–05:22] Anthropic was founded in 2021 by former OpenAI employees prioritizing ethical guardrails. Their CEO, Dario Amade, is portrayed as quirky and committed to safety, nicknamed “Professor Panda.”
• Quote: “One way to think about Anthropic is that it’s a little bit trying to put bumpers or guardrails on that experiment.” — Amrit Ramkumar (04:59)
• [05:54] Anthropic is unique among AI companies for advocating oversight, even when contrary to its business interests, notably lobbying for regulation during both the Biden and Trump presidencies.

3. The $200 Million Pentagon Contract

• [06:27–07:53] Despite its safety branding, Anthropic signed a lucrative contract with the US Military, providing access to its Claude chatbot model. This was seen as paradoxical given their principled stance.
• Amrit Ramkumar: “Military contracts are an enormous deal for AI companies...If they say, we like your AI better than someone else’s AI, that has immense value to shareholders.” (07:53)
• Anthropic’s unique status: Via partnership with Palantir, Claude is the only AI model cleared for classified defense work at present.

4. Ideological Clash with the Trump Administration

• [08:58–09:49] Almost as soon as the contract was signed, tensions rose. The administration labeled some AI models, Anthropic included, as “too woke.”
• Secretary of Defense Pete Hegseth’s blunt stance: “Department of War AI will not be woke. It will work for us. We’re building war ready weapons and systems, not chatbots for an Ivy League faculty lounge.” (01:14)
• Anthropic drew “red lines” in its terms: no AI for autonomous weapons or domestic surveillance.

5. The Venezuela Operation—Ethics in Action

• [11:55–13:00] In January, the US military used Claude in a strike operation in Venezuela where President Nicolas Maduro was captured, resulting in fatalities. This deployment triggered internal concern at Anthropic and increased scrutiny.
• Quote: “This is one of the first times we know that a specific model was used in an operation like this where people died. ...Afterward, people at Anthropic started asking some questions about how and why.” — Amrit Ramkumar (12:07)
• Anthropic staff sought clarity on Claude’s usage, prompting Pentagon alarm.

6. Breakdown in Trust: The Supply Chain Threat

• [13:43–16:03] The Pentagon began reviewing its partnership and threatened to label Anthropic a “supply chain risk”—a designation mainly reserved for companies with ties to adversaries. For an American company, this is an unprecedented and serious business threat.
• Quote: “If they’d go through with that, ...all Pentagon vendors and contractors would have to certify that they don’t use Anthropic’s models in their government work.” — Amrit Ramkumar (15:02)
• Such a move could devastate Anthropic’s government business and reputation.

7. High Stakes: What’s Next for Both Sides

• [16:03–17:46]
  • For Anthropic: Possible isolation from government/military, pressure to compromise on ethics.
  • For Pentagon: Losing exclusive access to Claude—only model cleared for classified uses—could pose risk to military capability.
• Quote: “A lot of very smart people say that would be counterproductive for US national security, for the goals of the administration. ...Cutting yourself off from some of the most advanced models would not be a great strategic choice.” — Amrit Ramkumar (16:50)

8. Broader Implications: The AI Arms Race

• [17:46–18:32]
• The dispute exemplifies how quickly AI is being adopted at the highest levels of government and how unresolved questions about regulation, ethics, and national security are coming to the fore.
• Quote: “It shows that AI adoption at the highest levels of government...is happening very quickly...the AI arms race and the geopolitical implications are only accelerating.” — Amrit Ramkumar (17:55)

---

Notable Quotes & Memorable Moments

• On Pentagon priorities:
  “We will not employ AI models that won’t allow you to fight wars.” — Pete Hegseth (13:00)
• On Anthropic’s mission:
  "They're the one big AI model developer that has been fighting the Trump administration on this issue and doesn't like the Trump administration's laissez faire approach to AI regulation." — Amrit Ramkumar (05:56)
• On the unique approval status of Claude:
  “Claude...is the only one that’s been approved to be used in classified scenarios. None of the other models have that approval yet, and it has already been embedded, so you can’t really strip that out.” — Amrit Ramkumar (16:50)

---

Important Timestamps

• 00:05–00:36 — Defense Summit intro & focus on Anthropic’s AI
• 04:02–05:22 — Anthropic’s foundation & commitment to safety
• 06:27–07:53 — The lucrative contract & why AI companies court the Pentagon
• 08:58–09:49 — Culture clash, "Woke AI" and Anthropic’s red lines
• 11:55–12:49 — Venezuela operation & internal questioning at Anthropic
• 13:43–15:02 — Threat of “supply chain risk” designation
• 16:03–17:46 — Analysis: What happens next for Anthropic and the Pentagon
• 17:46–18:32 — Conclusion: The AI arms race and its accelerating pace

---

Summary Takeaway

This episode navigates a tense intersection of business interests, ethical AI development, and political backlash, as Anthropic finds itself under fire from a defense community intent on maximum operational flexibility. The standoff raises fundamental questions about how powerful new technologies will be governed and used in high-stakes government contexts.

For listeners:
Expect a balanced blend of Wall Street Journal reporting, pointed primary-source quotations, and an exploration of issues that are likely to shape the future of war, technology, and democracy itself.

Transcript

[Jessica Mendoza]

Top names in the US Defense industry got together last week for a meeting in West Palm Beach. Here's our colleague Amrit Ramkumar.

[Amrit Ramkumar]

The Defense Technology Summit in Florida on Tuesday was really a gathering of some top Pentagon officials and people in the tech world. And the motivation is to embed the most advanced technology throughout the US Military.

[Jessica Mendoza]

These days, talking about tech means talking about AI. But there was one thing they were really buzzing about

[Amrit Ramkumar]

and that was the AI company Anthropic and the use of its tools in the military. And it's the latest sort of development in this saga that's consumed a lot of Washington.

[Jessica Mendoza]

This saga goes back years and has to do with the way Anthropic's AI is being used.

[Amrit Ramkumar]

They are clashing with the Pentagon over how the military use quad Anthropic's AI models and Anthropic is trying to put up some guardrails and prevent some uses. And they've been fighting over that for many months now.

[Jessica Mendoza]

One of the loudest voices in the disagreement is Secretary of Defense Pete Hegseth,

[Amrit Ramkumar]

Department of War AI will not be woke. It will work for us. We're building war ready weapons and systems, not chatbots for an Ivy League faculty lounge.

[Jessica Mendoza]

What is at stake here? Why does it matter if Anthropic and the government are on the outs?

[Amrit Ramkumar]

This fight is really about AI's use in the military and the government broadly. It's about whether the technology can be used for in the future, autonomous weapons potentially, or for surveillance and law enforcement. And that has far reaching consequences that will probably not be clear for a while. But this is the first salvo in what is going to be a very long, complicated battle.

[Jessica Mendoza]

Welcome to the Journal, our show about money, business and power. I'm Jessica Mendoza. It's Monday, February 23rd. Coming up on the show Anthropic vs. The Pentagon.

[Indeed Sponsor]

This episode is brought to you by indeed. Hiring isn't just about finding someone willing to take the job. It's about finding someone with the right skills and background who can move your business forward. And a good way to start your search is with INDEED Sponsored Jobs. It's one of the best ways to make your job post stand out and reach the candidates you're looking for faster. According to Indeed data, sponsored jobs posted directly on Indeed are 90% more likely to report a hire than non sponsored jobs. Plus there's no monthly subscriptions or long term contracts. You're only paying for results. Find the candidates who check all your boxes faster with INDEED Sponsored Jobs. Listeners of this show will get a $75 sponsored job credit to help get your job the premium status it deserves@ Indeed.com journal. That's Indeed.com journal right now and support the show by saying you heard about Indeed on this podcast. Indeed.com journal terms and conditions Applying Hiring do it the Right Way With Indeed

[Energy Sponsor]

behind every moment, there's a team working to make it happen. Deepening connections, strengthening the energy grid, and preparing for tomorrow. Most people never see that work, but they feel it every day. We are America's electric companies and our commitment is provide energy as affordably as possible, build a stronger tomorrow, and deliver for the people who depend on us. Learn [e-mail].

[Jessica Mendoza]

Anthropic was founded in 2021 after a group of former OpenAI employees decided to start an alternative they are a bunch

[Amrit Ramkumar]

of people who got uncomfortable with the idea that OpenAI was favoring technology advancement over responsible use of the technology.

[Jessica Mendoza]

OpenAI has said it prioritizes AI safety. Still, Anthropic's reputation grew.

[Amrit Ramkumar]

This is just like a quirky bunch of San Francisco characters. The CEO Dario is known as Professor Panda. He has a chicken coop in his backyard and he has a camera on them at all times. So really, yeah, they're sort of this ragtag bunch in some ways that came out of OpenAI and they've now suddenly become maybe one of the most important AI developers.

[Jessica Mendoza]

That founding group has talked a lot about the company's focus on safety. Here's CEO Dario Amade on 60 Minutes last year.

[Amrit Ramkumar]

One way to think about Anthropic is that it's a little bit trying to put bumpers or guardrails on that experiment. Right. Anthropic has tried to brand themselves as the AI good guys. They say AI could destroy the world and is too powerful and is improving too quickly. So the government needs to do all these regulations and slow it down.

[Jessica Mendoza]

The company supported oversight even if it didn't seem to be in their best interest. For instance, Anthropic has lobbied in favor of state led AI regulation. That philosophy was a good match with the Biden administration's priorities.

[Amrit Ramkumar]

The Biden administration really did focus on AI safety, transparency and taking steps to put some guardrails up. So when the Trump administration came in and on day one basically said we're getting rid of all of that, Anthropic fought back and they've basically spoken publicly about the need for guardrails.

[Jessica Mendoza]

Here's Dario Amade again.

[Amrit Ramkumar]

I think we do need to be raising the alarm. I think we do need to be concerned about it. I think policymakers do need to worry about it. They're the one big AI model developer that has been fighting the Trump administration on this issue and doesn't like the Trump administration's laissez faire approach to AI regulation. They don't like exporting advanced AI chips to China, and they speak their mind. The CEO, Dario Amade, is one of the only notable CEOs who has spoken out consistently against Trump in this second term.

[Jessica Mendoza]

At the same time, Anthropic was looking to grow its business and find new customers. And one of the most high value customers for any AI company is the US Military. Anthropic had already formed a partnership with Palantir, a data analytics company that has long worked with the Department of Defense and federal law enforcement. That partnership allowed Palantir to offer Claude to Palantir customers, which laid the groundwork for Anthropic's technology to be used in defense work. And in the summer of 2025, Anthropic landed a major $200 million contract with the US military. The idea was that defense operations could start using Anthropic's chatbot Claude to help with their work.

[Amrit Ramkumar]

The government is trying to embed AI everywhere it can, and Anthropic's Claude models are simply too good. They're often better at things like coding or other sorts of applications.

[Jessica Mendoza]

The government had also granted a special level of access to Anthropic through the company's partnership with Palantir, meaning Claude could be used in classified settings. It's the only AI model that currently has that level of clearance, and getting it takes a long time. So if Anthropic says it's really concerned with safety, it's really socially conscious, that sort of doesn't strike me as the type of place that would go for a military contract. Why did the company go down that route?

[Amrit Ramkumar]

Military contracts are an enormous deal for AI companies trying to get their foot in the door with the US Government. AI is seen as the future technology across the board, but it has so many military applications. People talk about autonomous drones using it to crunch numbers more efficiently. I mean, you can really embed it throughout your operations. It could be used down the road in battle. And more importantly for these companies, it's seen as just very lucrative because the military is seen as a huge, huge customer. And if they say, we like your AI better than someone else's AI, that has immense value to shareholders. So for all of those reasons, it's important

[Jessica Mendoza]

at the time. Anthropic noted in a press release that the partnership would dramatically improve intelligence analysis and enable officials in their decision making processes. How quickly does the culture clash between Anthropic and the Trump administration start to bubble up after this contract is signed?

[Amrit Ramkumar]

Almost immediately. Tension started running hot. Around the same time. It's important to note that the Trump administration signed an executive order about Woke AI that essentially implicitly was calling out Anthropic and others that they felt their models were too woke and if you asked it certain questions about how many genders there were or other things like that, you would get a woke response. And Anthropic was widely seen as a main target.

[Jessica Mendoza]

Dario Amade has said that the company isn't woke and that Anthropic doesn't have political motivations.

[Amrit Ramkumar]

And then right away, Anthropic and the military there were discussions about what Claude could and couldn't be used for. And in Anthropic's terms of service it states very clearly it cannot be used for anything related to domestic surveillance and autonomous weapons is another red line that the company has drawn.

[Jessica Mendoza]

Anthropic's usage guidelines prohibit Claude from being used to facilitate violence, develop weapons or conduct surveillance. An Anthropic spokesman said the company can't comment on whether Claude or any other AI model was used for any specific operation, classified or otherwise. He added that any use of Claude is, quote, required to comply with our usage policies which govern how Claude can be deployed. We work closely with our partners to ensure compliance. But Anthropic's stance was about to be tested. The United States just attacked Venezuela's capital, hitting military facilities in the country. President Trump ordered this operation and tensions between Anthropic and the Defense Department would spiral into a full blown standoff. That's next.

[Apple Card Sponsor]

Brought to you by Apple Card. Hey, you could be earning 2% daily cash back on that purchase and that one and even that one. That's because Apple card users earn 2% daily cash back on every purchase, including everyday items you buy online or in store when using their Apple Card. With Apple Pay, not an Apple Card customer, you can apply in the Wallet app on iPhone subject to credit approval. Apple Card issued Goldman Sachs Bank USA Salt Lake City branch terms and more at Apple Co Benefits say you've always

[Empower Sponsor]

wanted to sneak away for a last minute long weekend in Napa. Here's the thing. If you get smart with your money, you could do things like that. With Empower, you can start making the most out of your money so you can go out and live a little Isn't that why we work so hard to have some fun with our money? Like treating yourself to a private wellness retreat or scoring incredible seeds to a concert you'll never forget? So use Empower and get good at money so you can be a little bad. Join their 19 million customers today at empower.com, not an empower client, paid or sponsored.

[Jessica Mendoza]

In January, the US Military ordered a strike on Venezuela and captured its president, Nicolas Maduro. The Wall Street Journal has since reported that Claude was used in that operation.

[Amrit Ramkumar]

It's been very hard, as you can imagine, to unearth too many details about how these models get used, but we are told they offer some benefits, especially in the planning stages of these different missions. So, yes, this is one of the first times we know that a specific model was used in an operation like this where people died. And that was on a pretty large scale. Afterward, people at Anthropic started asking some questions about how and why and things of that nature. And then someone at Anthropic asked someone at Palantir how Claude was used in Venezuela.

[Jessica Mendoza]

An Anthropic spokesman said that these were, quote, routine discussions on strictly technical matters.

[Amrit Ramkumar]

And sort of the tone of the question and the substance set off alarm bells. And the Department of Defense found out,

[Jessica Mendoza]

and the Pentagon was quick to respond.

[Amrit Ramkumar]

Pete Hegseth, in a speech last month, he said essentially that we need AI tools that let us fight wars effective immediately. Responsible AI at the War Department means objectively, truthful AI capabilities employed securely and within the laws governing the activities of the department. We will not employ AI models that won't allow you to fight wars. And he was referring to Anthropic when he said that. We're told. So the Pentagon has sort of stayed at that spot and they're continuing to say, like, we need every AI model that we work with to agree to be used in all lawful use cases. And that has to apply across the board.

[Jessica Mendoza]

The administration started considering if it would cancel its $200 million contract with Anthropic, according to Wall Street Journal reporting. A spokesman for the Pentagon said that the Defense Department's relationship with Anthropic is under review. He said, quote, our nation requires that our partners be willing to help our war fighters win in any fight. A spokesperson for Anthropic said the company is, quote, committed to using Frontier AI in support of us. Tomorrow, Hegseth and Amadei are meeting in what's expected to be a tense discussion of how to move forward. And so where are we at in this feud? As the tension mounts. How bad has it gotten?

[Amrit Ramkumar]

It's gotten pretty bad in the sense that the Pentagon has now threatened to label Anthropic a supply chain risk, which sounds pretty vague but is actually a pretty big deal because that's something that's usually only used for companies associated with foreign adversaries.

[Jessica Mendoza]

The government typically labels a company as supply chain risk when it sees it as a potential threat to the economy or national security. The label means the company's products cannot be used for any government related work. And naming an American company a supply chain risk is extremely unusual.

[Amrit Ramkumar]

And if they'd go through with that, that would mean all Pentagon vendors and contractors would have to certify that they don't use Anthropic's models in their government work. So if you think about that, that means Anthropic's biggest investors and customers, they would all have to say in the context of their government work with the Pentagon, Anthropic and Claude had nothing to do with this. It's important to note too that Pete Hegseth, the Defense secretary combating Wokeism and woke companies, has been at the core of what he's tried to do in the military. And a lot of people just keep coming back to that ide that the administration feels Anthropic is too woke in different settings or too liberal. They can't quite shake that feeling.

[Jessica Mendoza]

But let's say the government does do that. If it follows through on this idea of labeling Anthropic supply chain risk. What would Anthropic's options be in that scenario?

[Amrit Ramkumar]

That's a really good question. I mean, it's safe to say Anthropic is in damage control mode. And if they do get designated a supply chain risk, extreme damage control mode to try to salvage something, because again, getting cut off from the military would be a huge deal for them. They've invested a lot in trying to become the AI company or one of the few focused on national security. And it would be a big blow because then you're potentially cut off from a huge customer. So it's, I mean, they would be in a real bind. They would have to maybe take other steps, like bending a knee a little bit in a way they haven't before. So that could put them in a very tough spot for sure.

[Jessica Mendoza]

What about the Pentagon? Are there any risks for the military if they stop working with Anthropic?

[Amrit Ramkumar]

A lot of very smart people say that would be counterproductive for US national security, for the goals of the administration, et cetera, et cetera. So they've made these threats, but a lot of people are looking at it and saying this would be sort of an own goal and wouldn't really benefit anyone. A lot of people look at this story and they think the Trump administration has Anthropic in a tough spot and is targeting them, and they have a lot of leverage. But we're told if the Trump administration decided to cut Anthropic off tomorrow and the military decided to stop using Claude altogether, there would be some real problems because Claude, again, is the only one that's been approved to be used in classified scenarios. None of the other models have that approval yet, and it has already been embedded, so you can't really strip that out. Cutting yourself off from some of the most advanced models would not be a great strategic choice.

[Jessica Mendoza]

And what does this say about the moment that we're in when it comes to AI adoption, particularly in the military, but in the government more generally?

[Amrit Ramkumar]

It shows that AI adoption at the highest levels of government and at the highest levels of companies is happening very quickly, and the government's going to be using it more and more. So it sort of supports that case that the sort of AI arms race and the geopolitical implications of all of that are only accelerating.

[Jessica Mendoza]

That's all for today. Monday, February 23 the Journal is a co production of Spotify and the Wall Street Journal. Additional reporting in this episode by Deborah Acosta, Vera Bergengrun, Keech Hagee, Berber Jin and Shalini Ramachandran. Thanks for listening. See you tomorrow.

[Energy Sponsor]

We all want to know what makes people tick.

[Indeed Sponsor]

How will they make decisions, handle feedback

[Energy Sponsor]

and work as a team? Team the Predictive Index behavioral assessment reveals how employees actually think, work and thrive. Like how Marcus fits in a role with structure. And Margo's strength is collaborating with others because when we understand each other, work just works. Take the behavioral assessment [e-mail] index.com.