The Defense Department has given Anthropic a Friday deadline to agree to its terms for military use of the company's AI technology, or it will force compliance through a Cold War-era law, according to a Pentagon official and a source familiar with the matter.
The dispute centers on how the Pentagon uses Anthropic's AI models, which it has had access to under a $200 million pilot program. During a Tuesday meeting with Defense Secretary Pete Hegseth, Anthropic CEO Dario Amodei walked through the company's red lines, which include autonomous weapons - where AI, not humans, makes final targeting decisions - and mass domestic surveillance of American citizens, the source said.
A Pentagon official said the dispute has nothing to do with those particular uses, adding that the Pentagon has always followed the law. "Legality is the Pentagon's responsibility as the end user," the official said.
Anthropic said in a statement that Amodei "expressed appreciation for the Department's work and thanked the Secretary for his service." The company said it "continued good-faith conversations about our usage policy to ensure Anthropic can continue to support the government's national security mission in line with what our models can reliably and responsibly do."
Axios was the first to report on the meeting.
The ultimatum
If Anthropic doesn't comply, Hegseth said he would invoke the Defense Production Act - a law that gives the president emergency authority to direct private companies to prioritize federal government orders - and force the company to allow its model to be used for military needs, the Pentagon official told ABC News. Both President Donald Trump and former President Joe Biden invoked the law during the COVID pandemic.
Hegseth also said he would label Anthropic a "supply chain risk," a penalty typically reserved for foreign adversaries that would halt the company's business with the Pentagon and its partners, the official added.
A source familiar with the meeting described it as respectful, with neither side raising voices, but it ended with the Pentagon making clear it would punish Anthropic if it didn't agree to its terms.
The pressure follows reporting that Anthropic and the Pentagon were at odds over continued use of Anthropic's AI on classified and unclassified systems. Pentagon spokesperson Sean Parnell said in a statement: "The Department of Defense's relationship with Anthropic is being reviewed. Our nation requires that our partners be willing to help our warfighters win in any fight. Ultimately, this is about our troops and the safety of the American people."
Elon Musk's xAI, Google, and OpenAI are also part of the Pentagon's pilot program. The Pentagon source said Grok, the AI tool developed by Musk's xAI, is on board with being used in a classified setting, while the rest of the companies are close.
Anthropic's defense stance
Anthropic has previously stated its AI models are not allowed to be used for weapons of mass destruction or decision-making that violates international human rights law. In the past, the company has said it will not sell its models to governments for surveillance that violates civil liberties.
The Wall Street Journal has reported that Anthropic's AI tool Claude was used in the U.S. military's capture of former Venezuelan President Nicolás Maduro, which included bombing several sites. An Anthropic spokesperson declined to comment: "We cannot comment on whether Claude, or any other AI model, was used for any specific operation, classified or otherwise."
The dispute sets up a test of how the Pentagon handles a major AI vendor that places limits on its technology. Hegseth's threat to use the Defense Production Act would mark the first time the law has been used against one of the Pentagon's commercial AI partners.
Why this matters for government professionals
For government employees working with AI systems, this is a dispute that will shape how federal agencies get their AI tools. The outcome will determine whether federal agencies can demand AI vendors follow rules beyond a company's stated terms, creating a test case for a broader procurement debate. A forced-compliance outcome could push more AI companies to restrict their government offerings or tighten usage policies, while a settlement could lead to new rules governing how AI tools are used in national security settings. Employees who follow this case will be watching for a decision that affects how procurement officers negotiate and which tools they can recommend for their operations.
Your membership also unlocks:
