WASHINGTON (AP) — Anthropic CEO Dario Amodei mentioned Thursday that the man made intelligence company “cannot in factual judgment of right and unsuitable accede” to the Pentagon’s calls for to permit unrestricted use of its expertise, deepening a public conflict with the Trump administration that’s threatening to pull its contract and rob other drastic steps by Friday.
The maker of the AI chatbot Claude mentioned in a press originate that it’s no longer walking away from negotiations but that original contract language got from the Protection Division “made virtually no progress on scuffling with Claude’s use for mass surveillance of People or in fully autonomous weapons.”
Sean Parnell, the Pentagon’s top spokesman, mentioned earlier on social media that the militia “has no hobby in the use of AI to conduct mass surveillance of People (which is illegal) nor will we desire to use AI to form autonomous weapons that operate with out human involvement.”
Anthropic’s policies prevent its models from being aged for those functions. It’s the final of its peers — the Pentagon additionally has contracts with Google, OpenAI and Elon Musk’s xAI — to no longer provide its expertise to a brand original U.S. militia inner network.
“It is a ways the Division’s prerogative to determine contractors most aligned with their imaginative and prescient,” Amodei wrote in a press originate. “But given the gigantic value that Anthropic’s expertise presents to our militia, we hope they rethink.”
Protection Secretary Pete Hegseth speaks at some stage in a cupboard meeting at the White Dwelling, Thursday, Jan. 29, 2026, in Washington. (AP Photo/Evan Vucci)
Protection Secretary Pete Hegseth speaks at some stage in a cupboard meeting at the White Dwelling, Thursday, Jan. 29, 2026, in Washington. (AP Photo/Evan Vucci)
Protection Secretary Pete Hegseth gave Anthropic an ultimatum on Tuesday after meeting with Amodei: Allow the Pentagon to use the company’s AI because it sees fit by Friday or probability losing its executive contract. Protection power officers warned that they may possibly maybe well waddle even extra and designate the company as a provide chain probability or invoke a Chilly Battle-period legislation known as the Protection Production Act to provide the militia extra sweeping authority to use its merchandise.
Amodei mentioned Thursday that “those latter two threats are inherently contradictory: one labels us a security probability; the opposite labels Claude as needed to national security.”
In a post earlier than Amodei’s announcement, Parnell reiterated that the Pentagon desires to “ use Anthropic’s model for all correct functions” but didn’t provide particulars on what that entailed. He mentioned opening up use of the expertise would prevent the company from “jeopardizing predominant militia operations.”
“We will not let ANY company dictate the phrases regarding how we create operational decisions,” he mentioned.
Emil Michael, protection undersecretary for analysis and engineering, later lashed out at the Anthropic CEO, alleging on X that Amodei “has a God-complex” and “desires nothing extra than to study out to in my belief succor watch over the US Protection power and is adequate inserting our nation’s security at probability.”
The talks that escalated this week began months prior to now. Amodei mentioned that if the Pentagon doesn’t rethink its repute, Anthropic “will work to enable a tender transition to 1 more provider.”
Sen. Thom Tillis, a North Carolina Republican who’s no longer searching for reelection, mentioned the Pentagon has been handling the topic unprofessionally while Anthropic is “attempting to attain their most efficient to relief us from ourselves.”
“Why in the hell are we having this dialogue in public?” Tillis informed journalists. “Here is no longer the capability you’re taking care of a strategic vendor that has contracts.”
He added, “When a company is resisting a market replacement for trouble of unfavorable consequences, you may possibly maybe accumulate to hearken to them after which in the back of closed doors decide out what they’re in actuality attempting to resolve.”
Sen. Mark Warner of Virginia, the ranking Democrat on the Senate Intelligence Committee, mentioned he modified into “deeply disturbed” by stories that the Pentagon is “working to bully a main U.S. company.”
“Sadly, this is extra indication that the Division of Protection seeks to completely ignore AI governance,” Warner mentioned in a press originate. It “extra underscores the need for Congress to attain sturdy, binding AI governance mechanisms for national security contexts.”
Protection Secretary Pete Hegseth stands outdoors the Pentagon at some stage in a welcome ceremony for the Japanese protection minister at the Pentagon in Washington, Jan. 15, 2026. (AP Photo/Kevin Wolf, File)
Protection Secretary Pete Hegseth stands outdoors the Pentagon at some stage in a welcome ceremony for the Japanese protection minister at the Pentagon in Washington, Jan. 15, 2026. (AP Photo/Kevin Wolf, File)
Whereas Pentagon officers dispute they continuously will apply the legislation with their use of AI models, the division has taken steps to substitute the tradition among the militia honest correct ranks.
Hegseth informed Fox News final February, weeks after turning into protection secretary, that “in the extinguish, we desire legal professionals who give sound constitutional suggestion and don’t exist to strive to be roadblocks to one thing else.”
The identical month, Hegseth additionally fired the quit legal professionals for the Army and the Air Pressure with out explanation. The Navy’s top lawyer had resigned quickly after the election in gradual 2024.
___
O’Brien reported from Providence, Rhode Island. Associated Press author Ben Finley contributed to this picture.
