OpenAI reached an agreement to use AI in classified systems of the US military

OpenAI’s agreement with the Pentagon comes shortly after President Donald Trump directed all federal government agencies to stop using Anthropic’s AI tools and US Secretary of Defense Pete Hegseth announced that the company would be considered a “supply chain threat” – a designation usually reserved for businesses from hostile countries.

Previously, Anthropic refused to make concessions in negotiations with the Pentagon on safety limits when using AI systems in automatic weapons and mass surveillance of American citizens. The US Department of Defense objected, asserting that it always operates within the framework of the law and that contract suppliers cannot set their own terms on how the military uses the product.

However, the new statement from Sam Altman, CEO of OpenAI, shows that the Pentagon has agreed to the same safety limits required by Anthropic. He wrote on

Altman said that OpenAI will develop technical safety measures to help AI models operate appropriately, and this is also what the Department of Defense wants. Additionally, the company will send engineers to the Pentagon to ensure the safety of the models.

“We are asking the Department of Defense to offer similar terms to all AI companies, and in our view, everyone should be willing to accept. We have expressed a strong desire to see the situation de-escalate, avoid legal and government action, and move toward reasonable agreements,” Altman said.

It is unclear why the Pentagon agreed to cooperate with OpenAI and how the new agreement differs from what Anthropic wants.

Theo ReutersAnthropic last year reached an agreement worth a maximum of 200 million USD to develop AI products for the US military. They are the first AI lab to deploy a model on the Department of Defense’s classified network, and the first company to develop a custom-built model at the request of a customer specializing in national security.

Anthropic attempted to discuss further terms of the contract before negotiations broke down. The company said it was “deeply saddened” to be labeled a “supply chain threat” by the Pentagon and planned to challenge the decision in court.

Theo TechCrunchmany employees from rival companies support Anthropic in its AI war with the Pentagon. An open letter titled We Will Not Be Divided calling on technology leaders to unite and take a stand has now received signatures from more than 600 Google employees and more than 90 OpenAI employees.