Pentagon Partners with Seven AI Firms to Deploy Classified‑Level Tools

TL;DR: The Pentagon announced agreements with seven AI companies to bring their tools onto classified networks, including Google’s Gemini 3.1 Pro model on the GenAI.mil platform. Officials say the move secures U.S. AI leadership and avoids reliance on a single vendor.

On Friday the Defense Department disclosed contracts with seven artificial‑intelligence firms to operate on its most secure networks. The companies are NVIDIA, OpenAI, SpaceX, Reflection, Amazon Web Services, Microsoft and Google. Their tools will reside on Impact Level 6 and 7 environments, the Defense Department’s highest security tiers for classified data.

The goal is to streamline data synthesis, sharpen warfighter decision‑making and raise situational awareness across the joint force. Officials said the arrangement also builds an architecture that prevents reliance on a single AI vendor and preserves long‑term flexibility.

The War Department and its partners asserted that American leadership in AI is indispensable to national security. They added that a thriving domestic ecosystem of model developers enables the full use of those capabilities in support of department missions. As mandated by President Donald Trump and Secretary Pete Hegseth, the department will continue to envelop warfighters with advanced AI to meet emerging threats.

In late April Google released its Gemini 3.1 Pro model on the Pentagon’s GenAI.mil platform for various defense use cases. GenAI.mil serves as the central AI hub where approved models can be accessed by authorized personnel.

The announcement follows a February dispute with Anthropic, which declined to allow its products to be used for autonomous weapons or surveillance of Americans. The Pentagon labeled Anthropic a supply‑chain risk, prompting Trump to order federal agencies to offload its tools, though a judge has since issued an injunction against that directive.

By spreading contracts across seven vendors, the Pentagon reduces the chance that any single company’s limitations or geopolitical pressures could disrupt critical AI access. The diverse toolbox also lets units pick models best suited for specific tasks, from language processing to image analysis.

Maintaining access to cutting‑edge commercial AI helps the joint force keep pace with adversaries who are likewise investing in machine‑learning for intelligence and targeting. At the same time, the arrangement must comply with existing procurement rules and oversight mechanisms that govern classified systems.

Observers will note how quickly the new tools move from pilot phases to routine use in training exercises and operational planning. Any delays or security hiccups could prompt revisions to the vendor list or platform policies.

Watch for upcoming after‑action reports from joint exercises that detail performance metrics and user feedback on the Gemini 3.1 Pro and other models now live on GenAI.mil.