Anthropic CEO Dario Amodei could still be trying to make a deal with Pentagon

1 month ago 17
Anthropic Co-Founder & CEO Dario Amodei speaks onstage during TechCrunch Disrupt 2023 astatine  Moscone Center.Image Credits:Kimberly White/Getty Images for TechCrunch

8:45 AM PST · March 5, 2026

Anthropic’s $200 cardinal declaration with the Department of Defense (DoD) broke down past week aft the 2 parties failed to travel to an statement implicit the grade to which the subject could get unrestricted access to Anthropic’s AI.

When the DoD made a woody with OpenAI instead, it seemed that the military’s narration with Anthropic would travel to a adjacent — but caller reporting from the Financial Times and Bloomberg accidental that Amodei resumed negotiations with Pentagon authoritative Emil Michael.

These talks are reportedly portion of an effort to compromise connected a declaration that outlines however the Pentagon tin proceed to entree Anthropic’s AI models.

It would beryllium a astonishment to spot Anthropic eek retired a caller deal, fixed however overmuch vitriol has been exchanged among the parties involved. But a compromise could inactive clasp entreaty for some sides — the Pentagon already relies connected Anthropic’s technology, and an abrupt power to OpenAI’s systems would beryllium disruptive.

The quality began erstwhile Anthropic CEO Dario Amodei voiced interest implicit a clause which allowed the subject to usage Anthropic’s AI for immoderate lawful use. Amodei asserted that the institution would not let for its exertion to beryllium utilized for home wide surveillance oregon autonomous weaponry, and wanted the declaration to much intelligibly prohibit those uses. When Anthropic refused to comply, the section turned astir and struck a deal with OpenAI instead.

Since then, figures connected some sides person been unfastened astir their frustrations. Michael called Amodei a “liar” with a “God complex.” Amodei threw immoderate jabs of his ain astatine the DoD and OpenAI CEO Sam Altman successful a message reportedly sent to Anthropic unit this week, calling the OpenAI woody “safety theater” and the messaging astir it “straight up lies.”

“The main crushed [OpenAI] accepted [the DoD’s deal] and we did not is that they cared astir placating employees, and we really cared astir preventing abuses,” Amodei wrote successful the memo.

Techcrunch event

San Francisco, CA | October 13-15, 2026

Defense Secretary Pete Hegseth has pledged state Anthropic a “supply concatenation risk,” fundamentally blacklisting the institution from moving with immoderate different institution that has immoderate concern with the U.S. subject — though helium has yet to instrumentality immoderate ineligible enactment to that effect. This benignant of designation is typically reserved for overseas adversaries, and it’s unclear whether it would past a tribunal challenge.

Amanda Silberling is simply a elder writer astatine TechCrunch covering the intersection of exertion and culture. She has besides written for publications similar Polygon, MTV, the Kenyon Review, NPR, and Business Insider. She is the co-host of Wow If True, a podcast astir net culture, with subject fabrication writer Isabel J. Kim. Prior to joining TechCrunch, she worked arsenic a grassroots organizer, depository educator, and movie festival coordinator. She holds a B.A. successful English from the University of Pennsylvania and served arsenic a Princeton successful Asia Fellow successful Laos.

You tin interaction oregon verify outreach from Amanda by emailing amanda@techcrunch.com or via encrypted connection astatine @amanda.100 connected Signal.

Read Entire Article