The U.S. War Department on Friday signed AI cooperation agreements with seven of America's largest technology and infrastructure companies to bring advanced AI models into top-secret network systems.
These contracts include SpaceX, OpenAI, Google, NVIDIA, Reflection AI, Microsoft, and Amazon Web Services. The agreements allow these companies' AI systems to be used in Level 6 (IL6) and Level 7 (IL7) secure environments for any legitimate operational purpose.
Inside the War Department's AI Deals
The War Department's Chief Technology Officer announced the agreement on May 1, 2024, stating that it represents a new step in building a War Department that is "AI-first," as officials have previously stated. The IL6 and IL7 environments are classified for secret and top-secret missions, meaning these AI models will operate alongside intelligence data and sensitive information.
“This is the latest step in the effort to create a PREFERENTIAL WAR DEPARTMENT OF AI,” the official account of the Office of the Undersecretary of War for Research and Technology stated .
Officials also Chia that collaborating with multiple vendors was intentional, aiming to avoid dependence on a single provider and to have the flexibility to choose between closed-source or open-source models.
NVIDIA will provide the open-source Nemotron suite , while Reflection AI – an NVIDIA-backed startup founded by former Google DeepMind engineers – will add several other open-source AI systems.
Google will provide the Gemini suite for all legitimate government purposes, while SpaceX plans to incorporate it into infrastructure related to its xAI Grok model.
Microsoft and AWS remain the pillars of cloud and infrastructure for deploying AI solutions on a large scale.
Internal usage has surged. The Department of War's GenAI.mil platform has attracted over 1.3 million users and tens of millions of usage commands in just five months since its launch, according to an announcement on May 1, 2024.
Anthropic did not participate due to conflicts over safety limits.
This list does not include Anthropic. Defense Secretary Pete Hegseth XEM Anthropic as a supply chain risk in February 2024, after Anthropic refused to lift restrictions on autonomous weapons and mass surveillance of its citizens.
"We will not let any single company dictate how we make operational decisions," War Department spokesman Sean Parnell stressed .
Subsequently, a federal judge temporarily blocked the ban, and the legal battle continues .
OpenAI chose a more limited scope of collaboration compared to its competitors. They stated that the agreement with the Department of War ensures three main commitments:
- Their AI models are not used for mass surveillance of the population.
- Not to be used to control automatic weapons, and
- Always maintain the established safety measures.
Other companies accept a broader term of “for any lawful purpose” without the publicly disclosed exceptions mentioned above.
The open-source trend will shape the next phase.
These agreements are part of the Department of War's AI Advancement Strategy, announced in early 2026, which focuses on an open architecture and open source code for combat, intelligence, and agency operations.
War Department officials said priority would be given to domestic suppliers, selecting transparent open-source solutions and promoting rapid testing, rather than relying solely on closed-code models.
The next step will be to XEM which models are approved for deployment at the IL6 level first and whether the security principles published by OpenAI are actually upheld as these systems scale up to highly confidential tasks.
