ARLINGTON, Va. — The Pentagon announced Friday that it reached agreements with major artificial intelligence companies to deploy tools on classified military networks while excluding Anthropic from the new rollout. The move is intended to speed secure AI adoption, broaden vendor options and reduce dependence on any single model provider, the War Department said in an official release May 1, 2026.
The department’s current list includes SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, Amazon Web Services and Oracle. Oracle’s inclusion followed the first wave of reporting, with Federal News Network reporting that the company was added after the initial Friday morning announcement, bringing the total number of participating firms to eight.
The agreements are designed to move frontier AI capabilities into Impact Level 6 and Impact Level 7 environments, the classified network tiers used for secret and highly restricted data. DefenseScoop reported that the companies signed formal agreements to deploy AI capabilities for “lawful operational use,” a phrase that has become central to the Pentagon’s dispute with Anthropic.
What the Pentagon AI deals include
The new agreements move the Pentagon beyond a narrow set of AI providers and toward a larger classified ecosystem. Officials said the tools could support warfighting, intelligence and enterprise operations by improving data synthesis, situational understanding and decision-making in complex environments.
The potential uses extend beyond chatbots. The Associated Press reported that military AI can help with target identification, weapons maintenance and supply lines, while also raising concerns about privacy, automation bias and the level of human oversight required in battlefield decisions.
The timing also reflects urgency inside the department. Reuters reported that the Pentagon has accelerated the process for bringing newer AI vendors onto secret and top-secret environments from 18 months or longer to less than three months, even as some military users have been reluctant to give up Anthropic’s tools.
Anthropic’s exclusion marks a sharp reversal
Anthropic’s absence is striking because the company was part of the Pentagon’s earlier frontier AI push. In August 2023, the Defense Department launched Task Force Lima to study generative AI and large language models across military use cases, saying the effort would balance adoption with risk management through the generative AI task force.
By July 2025, the Chief Digital and Artificial Intelligence Office had moved from study to procurement, announcing $200 million-ceiling awards to Anthropic, Google, OpenAI and xAI for agentic AI workflows across mission areas in a frontier AI contract announcement. OpenAI had already created a government-focused track, saying its first partnership under that effort would be a $200 million-ceiling Pentagon pilot through CDAO in its OpenAI for Government launch.
The relationship with Anthropic later broke down over the company’s requested limits on use of Claude. In March, Anthropic CEO Dario Amodei said the company’s concerns were its “exceptions on fully autonomous weapons and mass domestic surveillance,” while also saying Anthropic would support a transition for national security users where permitted in a statement on the dispute.
Why the shift matters
The Pentagon AI deals signal that the department is trying to avoid both technical lock-in and policy lock-in. Bringing in multiple providers could give commanders and analysts more options, but it also raises hard questions about model testing, data controls, audit trails, human review and accountability when AI is used in classified environments.
For Anthropic, the exclusion shows that ethical limits can shape access to government networks as much as technical performance. For the Pentagon, the next test will be whether a broader vendor mix can deliver secure, reliable tools quickly without creating new oversight gaps in intelligence and military operations.
