War Department Taps Seven Firms for AI on Classified Networks

The U.S. Department of War on May 1 signed agreements with SpaceX, OpenAI, Google, NVIDIA, Reflection AI, Microsoft and Amazon Web Services to run frontier artificial-intelligence models on Impact Level 6 and Impact Level 7 networks. The Department’s chief technology officer announced the package and the Office of the Under Secretary of War for Research and Engineering described the effort as creating an “AI-FIRST WAR DEPARTMENT.” The authorizations allow the systems to operate alongside secret and top-secret workloads for lawful operational use.

Officials said the contracts are intended to broaden vendor options and reduce dependence on a single provider. The agreements include both closed-source and open-weight models to support a range of missions and faster prototyping. The deals are folded into the Department’s AI Acceleration Strategy published earlier in 2026, which favors modular, open-source architectures, domestic suppliers and rapid experimentation across warfighting, intelligence and enterprise functions.

Under the agreements, NVIDIA will offer its open-source Nemotron family and Reflection AI will supply additional open-weight models. Google will make its Gemini family available and SpaceX is expected to contribute infrastructure connected to xAI’s Grok models. Microsoft and Amazon Web Services will remain the core cloud and infrastructure providers for the rollout.

OpenAI agreed to deploy its models on the classified networks while maintaining three public limits: the models will not be used for mass domestic surveillance, will not direct autonomous weapons, and will retain safety guardrails. Other contracted firms accepted broader “any lawful purpose” language for government use.

The Department reported that its GenAI.mil platform exceeded 1.3 million users and processed tens of millions of prompts within five months of launch.

Anthropic is not included after months of disagreement over guardrails and usage restrictions. Defense Secretary Pete Hegseth characterized Anthropic as a supply-chain risk after the company declined to lift limits on autonomous lethal weapons and mass domestic surveillance. Defense Department spokesman Sean Parnell wrote that the Department “will not let ANY company dictate the terms regarding how we make operational decisions.” A federal judge later blocked enforcement of the Department’s ban on Anthropic, and the legal matter remains active.

Officials said near-term milestones include which models clear IL6 certification first and how published safety commitments perform as classified workflows scale.