Antropic PBC said it would legally challenge the Pentagon’s decision to declare it a threat to the U.S. supply chain under powers normally granted to foreign adversaries, escalating a showdown with the Trump administration over artificial intelligence safeguards.
“We do not believe this action is legally sound, and we see no other option but to challenge it in court,” CEO Dario Amodei said in a blog post Thursday, hours after Bloomberg News reported the Pentagon’s findings.
Amodei said in the post that defense officials notified the company of the supply chain designation on Wednesday. The move could jeopardize the company’s $200 million contract to provide sensitive AI tools to the Department of Defense and prevent Anthropic from partnering with other companies in defense work.
The decision was the culmination of weeks of tense negotiations between Amodei and government officials over the U.S. military’s access to Anthropic’s technology. Talks broke down last week after the company demanded assurances that its AI would not be used for mass surveillance of Americans or to deploy autonomous weapons, and Defense Secretary Pete Hegseth threatened the company with a supply chain risk designation.
Although Anthropic still intends to contest the move, Amodei said the statute that was invoked (Section 3252 of the U.S. Code, which governs the military) is tailored narrowly enough so as not to affect other Anthropic operations unrelated to a specific Department of Defense contract.
This comes as some relief to customers and investors who were worried that the company would be unable to do business with companies that worked with the Department of Defense. A Microsoft spokesperson said Thursday that the company has concluded that it can continue working with Anthropic on non-defense projects.
Still, the designation means the company must stop working with another military contractor, Palantir Technologies. This includes Palantir’s use of Anthropic’s Claude in a digital mission control platform known as the Maven Smart System that was deployed in the U.S. military’s Iran operations.
The Pentagon’s decision also threatens to slow broader efforts to accelerate AI adoption across the U.S. military. Until recently, Anthropic offered the only AI system capable of operating in the Department of Defense’s classified cloud, and its Claude Gov tool has become a preferred choice among defense personnel due to its ease of use.
“This is a great feature,” said Lauren Kahn, a senior research analyst at Georgetown University’s Center for Security and Emerging Technologies, who said that removing it “would be painful for everyone involved.”
Defense officials said earlier Thursday that the decision was “effective immediately,” but Anthropic’s Claude AI tools are still actively used in the U.S. military’s Iran operations, according to people familiar with the matter. In his Feb. 27 warning to the company, Hegseth outlined a six-month transition period for moving AI operations to other providers.
Mr. Hegseth informed Congress of his decision in a letter to the Republican and Democratic leaders of the House and Senate Armed Services, Appropriations, and Intelligence Committees, according to correspondence obtained by Bloomberg.
“This decision is based in part on a risk analysis by the DoW and input from senior DoW officials that restrictions on the use of covered entities’ products and services pose a national security risk to the DoW supply chain,” Hegseth wrote, referring to Anthropic and using the acronym for the Department of the Army, which he currently prefers for the Department of Defense.
Anthropic is also bracing for job losses at private agencies after President Trump last week asked the federal government to remove the company as a contractor. The Treasury Department and the General Services Administration have already announced their intention to cease doing business with the company.
Amodei said in a post Thursday that he has been in discussions with the Department of Defense in recent days, finding it “productive” about how to address the company’s concerns about its AI safeguards. He said Anthropic plans to continue providing products to the military as long as it is allowed to do so.
Emil Michael, the undersecretary of defense for research and engineering, who has been negotiating with Mr. Amodei over the past few weeks, said on XPost late Thursday that there would be no further discussions. “I would like to end all speculation. @DeptofWar negotiations with @AnthropicAI are not active,” he wrote.
Currently valued at $380 billion, Anthropic is on track to generate approximately $20 billion in annual revenue, and its operating rate has more than doubled since the end of last year. But a dispute with the Pentagon has clouded the company’s prospects.
Applying supply chain risk labels to U.S. companies is unprecedented and would go beyond the scope of the law, said Charlie Block, a senior fellow at the Institute for Law and AI.
“This is not an authority to bring down large American companies that have contractual disagreements with the U.S. government,” he said. “This is an agency that deals with things like espionage by Chinese companies.”
Amodei acknowledged the bitter nature of the meeting and apologized for comments he made in an internal memo that surfaced on Wednesday. In a memo reported by the Information, Amodei accused archrival OpenAI of acting opportunistically and sacrificing safeguards in the Pentagon agreement announced by CEO Sam Altman hours after Trump and Hegseth ordered the U.S. government to sever ties with Anthropic.
Amodei also wrote in the memo that he believes the real reason the administration dislikes Anthropic is because it doesn’t donate to Trump, support his AI policies, or give him “dictator-style praise.” On Thursday, Amodei said: “Anthropic did not leak this post or direct anyone else to leak it.”
“It’s not in our interest to escalate,” he said, adding that the article was written hours after Trump and Hegseth’s posts and the OpenAI announcement. “It’s been a tough day for the company. I’m sorry for the lack of posting.”
Eastlund, Munson and Turner write for Bloomberg.
