The agreements expand the Pentagon’s access to commercial AI, accelerating its multi-vendor strategy amid its dispute with Anthropic.

The U.S. Department of Defense (DoD) signed agreements with seven frontier AI developers to deploy their models on classified networks.

According to its official release, the DoD reached agreements with OpenAI, Microsoft, Amazon Web Services, NVIDIA, Oracle, SpaceX, and the lesser-known Reflection. The Pentagon will integrate the models into “the Department’s Impact Level 6 (IL6) and Impact Level 7 (IL7) network environments [which] will streamline data synthesis, elevate situational understanding, and augment warfighter decision-making in complex operational environments.”  

The move comes on the heels of Google’s reported agreement with the DoD, which allows the use of the company’s AI systems for “any lawful government purpose.” Together, the agreements signal a rapid expansion in the government’s use of commercial AI technologies across classified systems.

The rapid expansion is part of the Pentagon’s broader AI Acceleration Strategy, which prioritizes faster deployment of advanced capabilities while reducing dependency on any single provider. The DoD’s partner roster now features all of the top-tier frontier AI developers, with the notable exception of Anthropic.

The absence of Anthropic follows an ongoing dispute between the company and the U.S. government over the use of its AI systems in military contexts. The disagreement, which began earlier this year, centered on restrictions Anthropic sought to maintain around surveillance and autonomous defense applications.

Despite the ongoing rift, Anthropic is still being used for classified government work. The company’s “Mythos” model remains under evaluation in controlled environments, where it is being tested for its ability to identify cybersecurity vulnerabilities and emerging threats.