The Pentagon has decided that Anthropic poses a supply-chain risk and wants everyone to abandon the outfit.
The defence department formally notified Anthropic’s leadership that the company and its AI tools present security threats, a senior Pentagon official said. The move is usually saved for foreign adversaries, and it capped a weeks-long row between the startup and the Pentagon.
The Pentagon wants Anthropic to allow AI use in all lawful-use cases, while Anthropic is pushing for explicit guarantees on autonomous weapons and mass domestic surveillance. The official said the military will not let a vendor wedge itself into the chain of command and put service members at risk by limiting lawful use of a critical capability.
The US Defence Department’s Secretary, Pete Hegseth, said last week he would issue the supply-chain risk notice, yet talks continued as both sides sought to reach a deal. Department of Defence, undersecretary of defence for research and engineering Emil Michael said, “I want to end all speculation: there is no active @DeptofWar negotiation with @AnthropicAI,” on X.
The fallout could spread to Anthropic partners and investors, including Lockheed Martin, Amazon.com and Alphabet’s Google. Some warn it is one of the first times this designation has been used on a US company, and it could chill other firms eyeing government work.
The supply-chain risk tag can be broad or narrowly tailored, and analysts are already arguing about how far the Pentagon can push it. Hegseth posted last week that any company working with the military cannot do business with Anthropic, which some analysts said goes beyond the Pentagon’s legal authority.
Those analysts argued the defence department can only bar use of Anthropic’s Claude inside Pentagon work, not ban all customer use if they hold defence contracts.
The move followed a report by The Information that Amodei sent staff a memo saying the administration is targeting Anthropic because it has not “given dictator-style praise” to President Trump and has resisted the administration’s AI agenda. The memo, viewed by The Wall Street Journal, took a swing at OpenAI for accepting a deal for its systems to be used in classified settings.
Anthropic chief executive Dario Amodei wrote that OpenAI chief executive Sam Altman was attempting “spin/gaslighting” while claiming to back Anthropic’s stance and rushing to sign its own pact. “It is working on some Twitter morons, which doesn’t matter, but my main worry is how to make sure it doesn’t work on OpenAI employees. Due to selection effects, they’re sort of a gullible bunch.”
He has since backed down on these comments, saying that he wrote the memo right after Trump and Hegseth announced measures that would severely impair Anthropic’s relationships, as OpenAI announced its Pentagon deal.
“It was a difficult day for the company, and I apologise for the tone of the post,” he wrote.
“It does not reflect my careful or considered views. It was also written six days ago, and is an out-of-date assessment of the current situation.” Amodei said.
Amodei said Anthropic believes the supply-chain risk designation is not legally tenable, citing the letter’s narrower scope. He said it supports Anthropic’s view that the designation “applies only to the use of Claude by customers as a direct part of contracts with the Department of War, not all use of Claude by customers who have such contracts,” despite Hegseth’s earlier, broader warning.
A Microsoft spokesperson said, “Our lawyers have studied the designation and have concluded that Anthropic products, including Claude, can remain available to our customers, other than the Department of War, through platforms such as M365, GitHub, and Microsoft’s AI Foundry and that we can continue to work with Anthropic on non-defence related projects.”
Founded by former OpenAI executives, Anthropic had been the only company approved for classified settings until xAI founder Elon Musk recently won approval. Anthropic’s Claude has been used in military operations in Iran and Venezuela during the operation to capture Venezuelan leader Nicolás Maduro, and it has a partnership with data-mining company Palantir Technologies, which supplies software to the defence department.
