Trump Directs the Government to Stop Using Anthropic’s AI

Summary

WSJ What’s News – PM Edition

Episode: Trump Directs the Government to Stop Using Anthropic’s AI
Date: February 27, 2026
Host: Alex Osola
Main Theme:
This episode centers on the Trump administration's abrupt order for all federal agencies, including the Department of Defense, to cease using Anthropic’s AI technology. The episode unpacks the escalating dispute between Anthropic and the U.S. military over the ethical guardrails on AI usage, as well as the broader implications for national security, business, and the future of AI in government.

---

Key Discussion Points and Insights

1. Escalating Military Preparations in the Middle East

• The U.S. is bolstering its military presence near Iran due to stalled nuclear negotiations and possible conflict.
• Nonessential American embassy staff are being withdrawn as tensions rise.
• President Trump stated:

  “We haven’t made a final decision. We’re not exactly happy with the way they’re negotiating. They cannot have nuclear weapons and we’re not thrilled with the way they’re negotiating. So we’ll see how it all works.” (01:35 – Trump)

2. Bill Clinton’s Congressional Testimony on Epstein

• Bill Clinton appeared before the House Oversight Committee, asserting his lack of knowledge about Epstein's crimes, and pledging not to "lie under oath."
• Democrats pressed for Trump to testify as well, citing his frequent association with Epstein in records.
• Representative Robert Garcia remarked:

  “He’s actually answering the questions fairly... He has not taken a pass on pleading the Fifth for any questions. He’s been very thorough.” (02:28 – Garcia)
  "We are now asking and demanding that President Trump officially come in and testify... He appears in the Epstein files... almost more than anybody else." (02:42 – Garcia)

3. Markets Update and Retail Shakeup

• Major U.S. indices dropped, with the Dow seeing the biggest single-day loss, especially hit by AI- and finance-related uncertainties.
• Reporter Jack Pitcher explained:

  “We saw tech and financials weighing on the stock indexes... When it comes to tech stocks, people are trying to figure out how some of the rapid AI developments we’re seeing are going to impact different business models." (03:38 – Pitcher)

• Bloomingdale’s thrives amid luxury retail sector turmoil by capitalizing on bankruptcies of rivals and rejuvenating its brand and in-store experiences.
• Suzanne Kapner noted:

  “They have laid in place a strategy to elevate Bloomingdale’s... and that includes renovating stores, adding new luxury brands like Christian Louboutin, and doing more immersive marketing events to make the shopping experience more like theater.” (04:40 – Kapner)

4. The Trump Administration’s Directive Against Anthropic’s AI

• President Trump publicly orders every U.S. federal agency to end use of Anthropic’s AI (Claude) within six months.
• Trump threatens civil and criminal consequences if Anthropic does not cooperate with the transition.
• This comes in response to Anthropic’s insistence on “red lines” concerning the military use of its AI—specifically, prohibiting usage in fully autonomous weapon systems (without a human in the loop) and for domestic mass surveillance.
• Trump administration is effectively canceling Anthropic’s $200 million Pentagon contract.
• Anthropic did not immediately comment, but previously stated it would “not back down from its guardrails.” (06:51 – Osola)

Military Use of AI – The Core Dispute

• Current Use: AI powers everything from document summarization to possible operational battlefield scenarios—recently, Claude was reportedly used in the operation to capture Nicolas Maduro.
• Dependency: The Pentagon relies on Claude, as it has been the only classified AI model available.
• Alternatives & Concerns: Elon Musk’s “Grok” is being rolled in, but is viewed as less reliable in sensitive government scenarios.

  “It would take many months to rip out Claude and substitute grok. And there are many things that Claude is better at than grok.” (08:18 – Ramkumar)

Anthropic’s Red Lines and Pentagon Pushback

• Anthropic’s stance:

  “…its technology can’t be used in fully autonomous weapons without a human in the loop or for domestic mass surveillance. Those are the two guardrails they want written in and codified in their contract with the DoD.” (08:55 – Ramkumar)

• Pentagon’s stance: Wants full freedom for all lawful use, asking Anthropic to trust government oversight.

Industry Context and Political Undertones

• OpenAI’s Position: Similar “red lines” as Anthropic, as affirmed by Sam Altman:

  “For all the differences I have with Anthropic, I mostly trust them as a company and I think they really do care about safety... I’ve been happy that they’ve been supporting our war fighters. I’m not sure where this is going to go.” (09:38 – Altman)

• Politics: Anthropic is labeled the “woke AI company” with Democratic donor ties, exacerbating tensions with the Trump administration.

  “A lot of this is about vibes, and Anthropic is seen as the woke AI company... They have these ties to organizations that are big donors to Democratic causes.” (10:12 – Ramkumar)

• Elon Musk/Xai: Willing to take a looser approach and cooperate more closely with the Pentagon, but questions remain about technical reliability.

---

Notable Quotes

• Amrith Ramkumar on Military Stakes of AI:

  “This is 100% about who controls how AI is used in the military. So those stakes are enormous because that’s going to be a topic that comes up again and again in the coming years and decades.” (00:38)

• Alex Osola on the Breaking Order:

  “Trump’s announcement appears to indicate that the administration is canceling Anthropic’s contract with the Pentagon, which is worth up to $200 million.” (06:51)

• Amrith Ramkumar on Dependency:

  “The military is somewhat dependent on Claude because it has been the only classified model available for many months now.” (08:18)

• On Pentagon’s Approval of Grok:

  “Officials within the National Security Agency said it’s not as reliable. If you test it in certain scenarios, when you’re talking about a military context or a high stakes government usage context, that reliability matters a lot.” (10:49 – Ramkumar)

---

Important Timestamps

• 00:18 – Key headlines rundown, including military buildup and the Trump-Anthropic dispute
• 01:35 – Trump speaks on Iran negotiations
• 02:28 – Bill Clinton congressional testimony; calls for Trump to testify
• 03:38 – Market and retail sector analysis
• 04:40 – Analysis of Bloomingdale’s success strategy
• 06:51 – Announcement of Trump’s order halting Anthropic AI use in government
• 07:41 – How military uses AI; dependence on Anthropic’s Claude
• 08:55 – Anthropic’s “red lines” and Pentagon’s position
• 09:38 – Sam Altman (OpenAI) comments on Pentagon-AI partnerships
• 10:38 – Political undertones and industry rivalries
• 11:30 – Discussion of Grok’s reliability and Musk’s Pentagon ties
• 11:40 – OpenAI’s massive fundraising; ongoing AI business developments

---

Other News Highlights

• OpenAI Funding: Raised $110 billion from SoftBank, Amazon, and Nvidia. Valuation prior to deal hit $730 billion.
• Neanderthal Study: New research reveals specifics of interbreeding between human ancestors and Neanderthals, with most encounters involving male Neanderthals and female humans.

---

Summary

This episode delivers a brisk yet deep dive into a historic clash between Anthropic and the U.S. government over the military’s ethical use of AI—a dispute that has immediate $200 million implications and long-term stakes for the direction of defense technology. It anchors this with context about military preparation, high-profile congressional hearings, financial markets’ AI concerns, and broader industry rivalries, all while illustrating how technical, legal, and political considerations are rapidly colliding in Washington.

Transcript

[Podcast Host]

Exchanges on the M and A and IPO landscape. Exchanges on the dynamics affecting global trade. For the sharpest analysis on finance, business and the economy, count on exchanges. The Goldman Sachs podcast Listen now.

[Alex Osola]

The US Is intensifying its military preparations in the Middle east for a possible attack on Iran. Plus, Bill Clinton tells Congress that he had no idea about Jeffrey Epstein's crimes. And Trump says he's ending the government use of Anthropic's AI. The administration's fight with Anthropic could have big national security implications.

[Amrith Ramkumar]

This is 100% about who controls how AI is used in the military. So those stakes are enormous because that's going to be a topic that comes up again and again in the coming years and decades.

[Alex Osola]

It's Friday, February 27th. I'm Alex Osola for the Wall Street Journal. This is the PM edition of what's news, the top headlines and business stories that move the world. The US Is gearing up for a possible war with Iran. The aircraft carrier USS Gerald R. Ford is approaching the Middle east today, adding a second carrier to the armada of American naval and air power already in the region. Talks between American and Iranian negotiators concluded yesterday with no deal on Tehran's nuclear and missile programs. The US Is pulling non essential staff from two of its embassies in the Middle east to prepare for possible counterstrikes from Iran. Speaking to reporters today, President Trump said the US And Iran weren't close to an agreement.

[President Trump]

Well, we haven't made a final decision. We're not exactly happy with the way they're negotiating. They cannot have nuclear weapons and we're not thrilled with the way they're negotiating. So we'll see how it all works.

[Alex Osola]

It's not clear when any attack would take place. Iranian and US Officials are supposed to meet in Vienna on Monday for talks. And in Chappaqua, New York, Bill Clinton testified today before the House Oversight Committee. The first time a former US President appeared appeared before a congressional committee under subpoena. Hillary Clinton testified yesterday. According to his prepared opening remarks for today's closed door session, Bill Clinton said he had no idea about Jeffrey Epstein's crimes. He also said that he was likely not to recall certain details and that he wouldn't lie under oath about events from two decades earlier. Representative Robert Garcia, a Democrat from California and the committee's ranking member, said Clinton was cooperative in his testimony.

[Representative Robert Garcia]

He's actually answering the questions fairly, I mean, to the best of his ability. There's no he has not taken a pass on pleading the Fifth for any questions. He's been very thorough.

[Alex Osola]

But Democrats also say Trump should testify, too.

[Representative Robert Garcia]

We are now asking and demanding that President Trump officially come in and testify in front of the Oversight Committee. He appears in the Epstein files next to next to Jeffrey Epstein and Ghislaine Maxwell almost more than anybody else.

[Alex Osola]

Representative James Comer, the Republican chair of the committee, said Trump has already answered questions about Epstein multiple times. Trump today said that he didn't like seeing Clinton deposed, but that quote, they certainly went after me a lot more than that. The three major US Indexes fell today. The Dow had the biggest loss, closing down 1.1%. The KBW Nasdaq bank index fell 4.9%, its largest one day drop since market turmoil last April related to tariffs. The country's largest banks tend to be a proxy for investors beliefs about the economy. Journal markets reporter Jack Pitcher says the reasons stocks dropped today aren't new.

[Jack Pitcher]

We saw tech and financials weighing on the stock indexes, which extended some storylines we've seen come up throughout the month. When it comes to tech stocks, people are trying to figure out how some of the rapid AI developments we're seeing are going to impact different business models. Meanwhile, in financials, there's been a lot of concerns about private credit and whether there's maybe some more risk in the system than people realized.

[Alex Osola]

For the month, The Nasdaq fell 3.4% while the S&P finished down 0.9%, both the sharpest losses since March 2025. The Dow ended up 0.2% in February, its 10th straight winning month. Bloomingdale's doesn't seem like a retailer that should be succeeding right now. Rivals like Saks and Lord and Taylor have filed for bankruptcy or gone out of business. But Bloomingdale's is thriving. It's posted five straight quarters of sales gains, and analysts expect that trend to continue. Suzanne Kapner covers retail for the Journal. Suzanne, why is Bloomingdale's doing so well when so many rivals have failed?

[Suzanne Kapner]

Well, the answer is really twofold. One, it is actually because rivals are failing that is giving Bloomingdale's an edge to win over consumers and brands. When Saks and its sister companies Neiman Marcus and Bergdorf Goodman filed for bankruptcy in January, that really opened up a lot of business that Bloomingdale's has been able to take advantage of. But it goes beyond that. They have laid in place a strategy to elevate Bloomingdale's, to rejuvenate it, to make it the exciting place to shop that it was during its heyday in the 1970s and 80s, and that includes renovating stores, adding new luxury brands like Christian Louboutin, and doing more immersive marketing events to make the shopping experience more like theater.

[Alex Osola]

Selling luxury brands is a key part of this strategy, but you write in your story that a decade ago many of those luxury brands wouldn't sell to Bloomingdale's. Why has that changed?

[Suzanne Kapner]

Well, in part because of the troubles at Saks, Neiman Marcus and Bergdorf Goodman. Those chains are sort of the true pinnacle of luxury department store retailing in the US So if they're already selling in Saxon Neiman Marcus, they probably don't need to be in Bloomingdale's. But now the dynamics have shifted and brands are a little more open to looking at alternative venues to distribute their products. And Bloomingdale's has really risen to the occasion. They are discounting less. They are giving the brands more freedom to express their vision in in store shops. They're sharing more data with the brands to try to create a true partnership.

[Alex Osola]

That was WSJ reporter Suzanne Kapner. Thanks Suzanne.

[Suzanne Kapner]

Thank you.

[Alex Osola]

Coming up, who decides what the military can do with AI? The fight has already begun. That's after the break.

[Twilio Advertiser]

This podcast is brought to you by Twilio. You bring the vision. Twilio brings the platform from real time messaging to AI powered insights. Twilio helps you build bold and personalized customer experiences. Twilio the ultimate toolbox for customer engagement. Learn [e-mail].

[Alex Osola]

President Trump has capped off weeks of tension between the Pentagon and Anthropic. He posted on Truth Social late this afternoon that he was ordering every federal agency in the United States government to stop using Anthropic's technology. He says the Department of Defense and other agencies using Anthropic's clawed models will have a six month phase out period. He said there would be civil and criminal consequences if the company was not helpful during the transition. Anthropic said earlier that it wouldn't back down from its guardrails around the military's use of Claude. Trump's announcement appears to indicate that the administration is canceling Anthropic's contract with the Pentagon, which is worth up to $200 million. Anthropic didn't immediately respond to a request for comment. I spoke earlier today with WSJ tech policy reporter Amrith Ramkumar. Amruth, let's start with some basics here. What does the Defense Department use AI for?

[Amrith Ramkumar]

The military is using AI in as many ways as it can. A few weeks ago, we broke the news that Claude was used in the Venezuela operation to capture former President Nicolas Maduro. And that was one of the first times that a lot of the public really saw that, wow, these tools are being used. It can be like summarizing documents. It could be back office functions. But you can also imagine a world where AI is controlling drone swarms. So the military is keeping its options open, essentially and saying we need this technology to be used in all lawful use cases as we see fit.

[Alex Osola]

And walk us through this dispute Anthropic is having with the Defense Department.

[Amrith Ramkumar]

The military is somewhat dependent on Claude because it has been the only classified model available for many months now. So the military recently approved Elon Musk's Grok for those scenarios. But it would take many months to rip out Claude and substitute grok. And there are many things that Claude is better at than grok. So on Tuesday, Pete Hegseth, the Defense Secretary, essentially threatened the anthropic CEO. He gave them until 5:01pm on Friday to do what the Department wants.

[Alex Osola]

A group of senators focused on defense have asked the two sides to reach a compromise. But what is it that Anthropic doesn't want Claude used for?

[Amrith Ramkumar]

Anthropic's red lines that it's very clear about are that its technology can't be used in fully autonomous weapons without a human in the loop or for domestic mass surveillance. Those are the two guardrails they want written in and codified in their contract with the Department of Defense. The Department is essentially saying, just agree to let us do all lawful things and we'll take care of your red lines and we'll abide by them. So in some ways, both sides are talking past each other. And we now know that OpenAI and Sam Altman, they've made clear that they want to maintain some of the same red wines.

[Alex Osola]

Right. Altman said that OpenAI was working on a possible deal with the Pentagon which could help solve the impasse with Anthropic. Here's what he said about the situation this morning on cnbc.

[Sam Altman]

Companies that choose to work with the Pentagon, as long as it is going to comply with legal protections and the sort of the few red lines that the field that we have I think we share with Anthropic and that other companies also independently agree with, I think it is important to. I've been for all the differences I have with Anthropic, I mostly trust them as a company and I think they really do care about safety and I've been happy that they've been supporting our war fighters. I'm not sure where this is going to go.

[Alex Osola]

Amrith, There have been tensions between the Trump administration and Anthropic for a while, right?

[Amrith Ramkumar]

A lot of this is about Vibes, and Anthropic is seen as the woke AI company. They've criticized the administration for giving AI chips to China. They have these ties to organizations that are big donors to democratic causes. They're big putting their own money into this organization that's essentially fighting most of the AI industry ahead of the midterm elections to support regulation of AI. So that broader context has a lot to do with this.

[Alex Osola]

You mentioned the Pentagon has also approved Xia's GROK tool for use in classified information recently. But you've reported that some government officials are warning against that use. Why is that?

[Amrith Ramkumar]

Over the last few years, officials within the gsa, the procurement arm of the federal government, they've raised concerns about grok. We've even seen some officials within the National Security Agency said it's not as reliable. If you test it in certain scenarios, when you're talking about a military context or a high stakes government usage context, that reliability matters a lot. So officials have raised those concerns. But the government has said they're comfortable with where GROK is now. And again, Elon Musk's relationship with the Trump administration and the Pentagon has a lot to do with this. He's willing to work with the military on AI usage and he's willing to take a looser approach than a lot of the other companies. So that definitely gave them.

[Alex Osola]

And we should note that Musk and Xai didn't respond to requests for comment. That was WSJ tech policy reporter Amrith Ramkumar. Thanks, Amrit.

[Amrith Ramkumar]

Thanks for having me.

[Alex Osola]

In other AI news, OpenAI says it's raised $110 billion in new funding. The financing includes $30 billion from SoftBank, a $50 billion commitment from Amazon, and $30 billion from Nvidia. The deal values the company at $730 billion before the investment. It suggests that investor interest remains Strong in the ChatGPT developer, which is expected to go public later this year. And finally, scientists know that our human ancestors interbred with Neanderthals, who thrived across Europe and Western Asia before going extinct thousands of years ago. Now new research gives us more detail about those sexual encounters. A study published yesterday says most of them were between male Neanderthals and female humans. The researchers figured this out by comparing the DNA from 73 modern women in Africa to genetic information from three female Neanderthals. And that's what's news for this week. Tomorrow you can look out for our weekly markets wrap up what's news in markets. Then on Sunday, we've got an episode of our USA250 series about the history and future of retirement in the US that's in what's New Sunday. And we'll be back with our regular show on Monday morning. Today's show is produced by Pierre Bienname with supervising producer Tali Arbel. Michael Lavalle wrote our theme music. Aisha El Musleam is our development producer, Chris Sinsley is our deputy editor and Falana Patterson is the Wall Street Journal's head of news audio. I'm Alex Osola. Thanks for listening. Foreign.

[Podcast Host]

Break free from the AI frenzy and truly transform your marketing. With Adobe, the path to ROI is clear and the opportunity is all around you. Let's turn AI's promise into your marketing reality. It starts with Adobe.