Pentagon signs AI agreements with six major vendors, expanding beyond Google

The Pentagon’s AI push just got much bigger — and much more concrete. On May 1, the Defense Department said seven companies will now be able to put AI products inside the military’s most sensitive network environments through GenAI.mil, its internal generative AI platform. That matters because this is not just “try some chatbots in the office” anymore. This is the point where commercial AI starts moving into secret and top-secret military workflows. (nextgov.com) ### Which companies got in? The list is broader than the early Pentagon AI lineup: Google, OpenAI, Nvidia, Amazon Web Services, Microsoft, SpaceX, and Reflection. The department said these companies were selected to support classified workflows tied to data synthesis, situational awareness, and faster decision-making for warfighters. The point is choice — not a single winner. (nextgov.com) ### What changed on May 1? The real step forward is access level. The Pentagon said these tools will be available in Impact Level 6 and Impact Level 7 environments. In plain English, that means secret and top-secret classified systems. So the news is not merely that the Pentagon likes AI vendors. It’s that it is now clearing a group of them to operate in the most sensitive parts of its computing stack. (nextgov.com) ### What is GenAI.mil, exactly? GenAI.mil is the Defense Department’s shared generative AI platform. The official pitch is simple: put leading AI models in the hands of military and civilian personnel “at all classification levels” so people can actually use them insid(nextgov.com)y and classified AI use. (ai.mil) ### Why does “multi-vendor” matter so much? Because the Pentagon is saying the quiet part out loud: it does not want vendor lock-in. The May 1 announcement said the architecture is meant to prevent dependence on one AI provider and preserve long-term flexibility for the Joint Force. (ai.mil)m does not freeze. (nextgov.com) ### Why is this bigger than Google? Until recently, the Pentagon’s frontier-model strategy was narrower. In July 2025, CDAO gave contracts with $200 million ceilings to four firms — Anthropic, Google, OpenAI, and xAI — to build agentic AI workflows for national securit(nextgov.com)ai.mil) ### What happened with Anthropic? Anthropic is the missing name, and that absence matters. In late February and March, a fight blew up over contract terms tied to autonomous weapons and domestic surveillance. Anthropic balked. OpenAI stepped in. The dispute tur(ai.mil)anies were accepting. (forbes.com) ### Why are Nvidia and SpaceX in this? Because the Pentagon is not buying only foundation models. Nvidia brings the chips-and-infrastructure layer that makes AI systems run. SpaceX points to communications, data movement, and operational integration. Reflection’s inclusion also suggest(forbes.com)t fits the whole architecture story — resilience through diversity. This last point is an inference from the vendor mix and the Pentagon’s anti-lock-in language. (nextgov.com) ### So what’s the bottom line? The Pentagon has moved from “which model should we test?” to “how do we run several of them inside classified operations without getting trapped by any one company?” That is the real story. The military is building an AI platform, not jus(nextgov.com)efield decision support. (nextgov.com)