The Department of Defense has finalised agreements with eight technology companies to deploy artificial-intelligence systems on its classified networks, expanding the use of frontier models in analysis, logistics and large-scale data processing inside the Pentagon's most sensitive enclaves.
OpenAI, Google, Microsoft, Amazon Web Services, Oracle, Nvidia, SpaceX and the smaller startup Reflection AI were among the firms approved. The department did not include Anthropic, citing what officials described as a supply-chain risk arising from a contract dispute earlier this year.
The Trump administration severed talks with Anthropic after the company refused to amend its acceptable-use policy to permit Department of Defense use of its Claude models for "all lawful purposes", including autonomous weapons and bulk surveillance. Anthropic has said its policies still allow a broad range of national-security applications.
Anthropic is now contesting the supply-chain designation in federal court. President Trump met with chief executive Dario Amodei at the White House on April 17 and afterwards told CNBC a deal with the company was "possible".
The eight cleared vendors will be allowed to operate in classified environments under arrangements that the department said would include human-in-the-loop controls for high-consequence tasks. Officials declined to disclose contract values but described the awards as multi-year.
Defence contractors and pure-play model developers have rushed to certify their systems for sensitive workloads since the administration loosened restrictions on military AI procurement last year. Industry analysts expect the agreements to generate several billion dollars of cumulative revenue across the awardees.
Civil-liberties groups criticised the awards, arguing that classified deployment will prevent independent assessment of model behaviour. The Pentagon said it would maintain its own internal red-team and evaluation programme.