Anthropic vs Pentagon: Hegseth sets Friday deadline for Claude AI use

The Pentagon has issued a final ultimatum to AI startup Anthropic, demanding unrestricted military access to its Claude model by Friday evening. Defense Secretary Pete Hegseth is pressuring CEO Dario Amodei to remove ethical guardrails that prevent the technology from being used in mass surveillance and autonomous weapons systems. If the company refuses to comply, officials have threatened to cancel its $200 million contract, label the firm a supply chain risk, and potentially invoke the Defense Production Act to force compliance. The conflict highlights a growing rift between the Trump administration's push for AI dominance and the safety-first philosophy of some Silicon Valley leaders. While competitors like OpenAI and Elon Musk’s xAI have already agreed to all lawful uses, Anthropic remains the sole holdout among the major defense-contracted AI firms.

[00:00] Daniel Brooks: From Neural Newscast, I am Daniel Brooks.
[00:03] Daniel Brooks: And I am Elise Moreau.
[00:05] Daniel Brooks: It is Thursday, February 26th.
[00:08] Daniel Brooks: The Pentagon is putting intense pressure on one of the most prominent AI startups in the country.
[00:14] Daniel Brooks: Defense officials have issued a final offer to Anthropic regarding the military use of its Claude software.
[00:23] Announcer: Defense Secretary Pete Hegseth wants full access to the model by tomorrow evening.
[00:29] Announcer: Today we will discuss the standoff over ethical guardrails and the threat of the Defense Production Act.
[00:35] Daniel Brooks: The situation reached ahead yesterday, February 25th, when Pentagon officials sent their best and final offer to the company.
[00:45] Daniel Brooks: It is a high-stakes moment for the $200 million contract signed last summer.
[00:51] Elise Moreau: Secretary Hegseth met with Anthropic CEO Dario Amadeh earlier this week to deliver the ultimatum.
[00:58] Elise Moreau: If the company does not grant all lawful use by Friday evening, they face the loss of all military business.
[01:05] Daniel Brooks: The Pentagon is even threatening to label Anthropic a supply chain risk.
[01:11] Daniel Brooks: That is essentially a government blacklist that would prevent them from working with any federal agency or partner in the future.
[01:19] Elise Moreau: This is a striking shift in tone for a company that was the only AI firm cleared for classified systems until recently.
[01:27] Elise Moreau: The government seems ready to sever a very deep technical relationship, Daniel.
[01:32] Daniel Brooks: The root of the disagreement is about how much control the military has over the software's internal rules.
[01:39] Daniel Brooks: Anthropic has built its reputation on being the most safety-forward firm in the industry.
[01:46] Elise Moreau: Amadeh is particularly concerned about mass surveillance and the use of AI in autonomous
[01:52] Elise Moreau: killing.
[01:52] Elise Moreau: He has been pushing for guardrails that would keep human judgment in the loop for lethal
[01:57] Elise Moreau: decisions.
[01:58] Daniel Brooks: The administration has been dismissive of these concerns.
[02:02] Daniel Brooks: Secretary Haigsef and other officials have labeled these safety measures as woke AI, suggesting
[02:09] Daniel Brooks: they are simply ideological roadblocks.
[02:13] Elise Moreau: The design of Claude is fundamentally built around these safety principles.
[02:18] Elise Moreau: If Anthropic agrees to remove them, it would change the architecture and the cultural identity
[02:23] Elise Moreau: of the product, Daniel.
[02:25] Daniel Brooks: Defense officials argue they are not asking for anything illegal.
[02:29] Daniel Brooks: They say they simply want a license to use the technology for all activities that are lawful under U.S. statutes.
[02:38] Elise Moreau: That is the exact language that competitors like OpenAI and Google have reportedly agreed to.
[02:44] Elise Moreau: Even Elon Musk's XAI chatbot was approved for classified use this past Monday after accepting those terms.
[02:51] Daniel Brooks: It leaves entropic as the lone holdout among the major players, Elise.
[02:56] Daniel Brooks: They're stuck between their ethical mission and the reality of doing business with the world's largest defense spender.
[03:02] Elise Moreau: There is also the looming threat of the Defense Production Act.
[03:07] Elise Moreau: This is a Korean War-era law that allows the president to force companies to prioritize government needs during national emergencies.
[03:15] Daniel Brooks: Invoking that act would be a major escalation.
[03:19] Daniel Brooks: It would essentially allow the military to seize control of the software and use it regardless of what the company's leadership thinks.
[03:27] Elise Moreau: It creates a difficult dilemma for a company planning to go public this year.
[03:32] Elise Moreau: Investors usually like government contracts, but they also value the unique safety brand Anthropic has built, Daniel.
[03:40] Daniel Brooks: Amadei has noted that revenue has actually grown while they have taken this stand.
[03:45] Daniel Brooks: But the pressure from the Pentagon is much more direct than any previous administration's requests.
[03:52] Elise Moreau: We have seen similar ethical debates in architecture when firms refuse to design prisons or border walls.
[03:59] Elise Moreau: This feels like a digital version of that conscientious objection.
[04:03] Daniel Brooks: Still, the Pentagon is focused on what they call the global AI arms race.
[04:09] Daniel Brooks: From their perspective, any guardrail that slows down a mission is a tactical disadvantage.
[04:15] Elise Moreau: We will see if a middle ground exists before the deadline tomorrow, February 27th.
[04:21] Elise Moreau: If not, the relationship between the Pentagon and one of its top AI partners could dissolve entirely by the weekend.
[04:28] Daniel Brooks: The implications for the broader industry are significant.
[04:32] Daniel Brooks: If the government can force an AI company to strip its safety features, it sets a massive precedent for every other firm in Silicon Valley.
[04:41] Elise Moreau: It is a question of who really owns the soul of a machine once it is sold to the state.
[04:46] Elise Moreau: We will be watching for Anthropics' response as that Friday deadline approaches.
[04:51] Elise Moreau: I am Daniel Brooks.
[04:53] Daniel Brooks: And I am Elise Moreau.
[04:55] Daniel Brooks: For more coverage, visit us online.
[04:57] Daniel Brooks: Neural Newscast is AI-assisted, human-reviewed.
[05:01] Daniel Brooks: View our AI transparency policy at neuralnewscast.com.
[05:06] Daniel Brooks: Neural Newscast uses artificial intelligence in content creation with human editorial review prior to publication.
[05:13] Daniel Brooks: While we strive for factual, unbiased reporting, AI-assisted content may occasionally contain errors.
[05:20] Daniel Brooks: Verify critical information with trusted sources.
[05:23] Daniel Brooks: Learn more at neuralnewscast.com.

Anthropic vs Pentagon: Hegseth sets Friday deadline for Claude AI use
Broadcast by