News

Pentagon Expands AI Agreements with Microsoft, Amazon, Nvidia, Oracle, and Others

• By John K. Waters
• 05/01/2026

The U.S. Defense Department has reached agreements with additional technology companies to expand the use of advanced artificial intelligence tools on classified military networks, according a statement issed by the Pentagon.

The agreements include Nvidia, Microsoft, Reflection AI, Amazon Web Services, and Oracle, and are intended to allow the Defense Department to use the companies’ AI tools for what the Pentagon described as “lawful operational use,” Bloomberg reported. The arrangements add to similar agreements involving OpenAI, Google, and SpaceX.

The new terms could give the Pentagon broader latitude to use AI systems in classified military operations, including secret combat-related work, while reflecting a widening split between defense officials and some AI companies over how far military use of advanced models should go.

The agreements follow a breakdown between the Pentagon and Anthropic, whose Claude model has been used in military systems. As we reported earlier, Anthropic sought restrictions on uses such as mass domestic surveillance of U.S. citizens and fully autonomous weapons systems. The Pentagon rejected those redlines and has since moved to line up other suppliers willing to accept broader terms.

The dispute has put renewed attention on how the U.S. military plans to use AI in classified environments. Defense Secretary Pete Hegseth told Congress on Thursday that the Pentagon follows the law and that humans make lethal decisions.

The Defense Department’s chief digital and AI officer, Cameron Stanley, said in a statement that the new agreements would support “human-machine teams” capable of processing large volumes of data, improving understanding in complex environments, and helping military users make decisions faster.

The Pentagon did not specify the exact operational uses planned for the new AI systems. The language in the agreements, including “lawful operational use,” gives the department more flexibility than the terms sought by Anthropic.

Amazon Web Services said it had supported the U.S. military for more than a decade and expected to continue working on modernization efforts, Bloomberg reported. Microsoft declined to comment, Nvidia did not immediately provide a comment, and representatives for Reflection AI and Oracle were not immediately available.

The Defense Department has been working since late 2025 to expand AI terms with major technology suppliers. Defense officials are also seeking to avoid dependence on a single AI vendor or one company’s restrictions, according to a Pentagon official briefed on the matter.

Nvidia’s agreement gives the Pentagon greater license than in previous AI arrangements, and Nvidia agreed not to impose usage policies or model licenses restricting the department’s use beyond what is required by U.S. law and constitutional authority.

The expansion comes as the Pentagon works to replace or reduce its reliance on Anthropic’s Claude, which has been used on the Maven Smart System, an AI-driven, Palantir-developed command-and-control platform for the U.S. Department of Defense that automates intelligence analysis. The platform has been used to support targeting and battlefield operations during Iran-related operations.

The broader adoption of AI in military systems has drawn concern from advocacy groups and AI safety critics, who warn that models can make errors, produce unreliable outputs, and encourage automation bias, in which human users place too much trust in machine-generated recommendations.

The Pentagon’s position is that AI will help military personnel process information faster while keeping humans responsible for decisions involving the use of force. The central policy question remains how much operational authority advanced AI systems should have in classified military settings, and what safeguards should be required before their use expands further.