What just happened? The Pentagon has finalized a sweeping set of AI agreements with some of the industry's biggest names, opening classified military environments to models and infrastructure from OpenAI, Google, Microsoft, Amazon, Oracle, Nvidia, SpaceX, and newcomer Reflection AI. The deals mark a significant expansion of how artificial intelligence is used inside the Defense Department, consolidating access that had previously been scattered or subject to tighter restrictions.

Defense officials have framed the effort as a necessary step to keep pace with both technological change and geopolitical competition. "We are equipping the warfighter with a suite of AI tools to maintain an unfair advantage and achieve absolute decision superiority," Emil Michael, the undersecretary of defense for research and engineering, told the Wall Street Journal.

Several companies in the group, including OpenAI and SpaceX, had already signed preliminary agreements with the military. The new contracts allow those relationships to expand into daily operations covering analysis, logistics, and large-scale data processing.

The vendor mix reflects the extent to which AI development is tied to cloud infrastructure. Microsoft, Amazon, and Oracle are not only building models but also providing the underlying compute environments where those models run. That combination allows the Pentagon to deploy AI systems within existing secure cloud frameworks, rather than building new infrastructure from scratch.

At the same time, the inclusion of Nvidia and Reflection AI signals a push toward open-source models. Nvidia's agreement centers on its Nemotron models, which are designed to support autonomous agents capable of carrying out multi-step tasks. Because these models are open, their internal architecture can be inspected and modified, which is useful for tailoring systems to specific defense use cases.

Nvidia CEO Jensen Huang has argued that this transparency can be an advantage in national security settings. "Safety and security is frankly enhanced with open-source," he said in a recent conversation with the Special Competitive Studies Project.

That view carries geopolitical weight. Chinese firms have aggressively distributed open AI models internationally, and US officials increasingly see domestic open-source development as a strategic imperative.

Reflection AI, a newer company backed by Nvidia, fits into that effort. Founded by former Google DeepMind researchers, the startup is working on open-source systems and has been involved in a government-supported initiative to develop models tailored for the South Korean market. The company has not yet released a model but is reportedly seeking funding at a $25 billion valuation.

"This shared understanding with the Pentagon is a first step in supporting US national security, and sets a precedent for how AI labs could work across the US government – from supporting our servicemembers to our scientists," a Reflection spokeswoman said.

The Pentagon's expanded roster of AI providers follows a breakdown in its relationship with Anthropic. The company's Claude models had been among the few available in classified settings, largely through Palantir's Maven platform. After a contract dispute, the Defense Department labeled Anthropic a supply-chain risk, effectively cutting off its systems and accelerating efforts to bring in other vendors. Defense Secretary Pete Hegseth, speaking during congressional testimony, called Anthropic CEO Dario Amodei an "ideological lunatic."

Anthropic is now challenging that designation in court. Its models had previously been used in military contexts, including during the Iran conflict and in an operation targeting Venezuelan leader Nicolas Maduro earlier this year.

But not everything is on the table. Several companies involved in the agreements have said their technologies will not be used for mass surveillance or autonomous weapons.

The Defense Department has said those uses would be illegal and has urged companies to trust its oversight. What is changing, though, is the role AI plays inside the military. With access to a broader range of models and the infrastructure to run them securely, the Pentagon is moving beyond experimentation.