This marks a first in American industrial history: a national company is being treated the same as the Chinese equipment maker Huawei.
This Thursday, Defense Secretary Pete Hegseth officially designated Anthropic as a “risk to the defense supply chain.” The decision takes effect immediately: the $200 million in military contracts are now void.
Financial consequences remain, for the moment, more limited than some had feared. The designation applies only to Claude’s use within direct contracts with the Pentagon. Microsoft, a major Anthropic partner, has already stated that its lawyers reviewed the text and concluded that Claude can be used in its commercial products (M365, GitHub, Azure AI Foundry) for all clients not tied to defense contracts. Amazon, a major investor in Anthropic, has not yet publicly commented.
End of the Contract with Palantir
By contrast, the rupture with Palantir is immediate. The Maven Smart System platform, which the U.S. military deploys, notably for its operations in Iran, and which relies on data streams built with Claude, will need a substitute. The Treasury and the General Services Administration have announced their intention to sever all commercial ties with Anthropic, in keeping with a direct directive from Donald Trump demanding the expunging of its solutions from all federal contracts.
Anthropic has stated that it will challenge the decision in court, arguing that the law (referred to as 10 USC 3252) requires the Secretary of War to employ the “least restrictive means necessary” to safeguard the supply chain. While awaiting the court’s ruling, the startup assures that it will continue to provide its models to the United States military “at nominal cost,” and with the support of its engineers, for as long as it remains legally permissible.
The Fallout from an Internal Memo
The crisis gained an additional political dimension with the leak of an internal memo by Dario Amodei, published by The Information.
Drafted on the day when Donald Trump ordered Anthropic’s exclusion from federal systems and when OpenAI announced at the last moment an agreement with the Pentagon, the note accused the government of sanctioning Anthropic for a lack of “dictator-style praise” toward the President.
The CEO issued a mea culpa: “It was a difficult day for the company, and I apologize for the tone of that message. It does not reflect my considered views.” He also clarified that Anthropic was not responsible for the leak.
