Pentagon: umowy z NVIDIA, Microsoft i AWS na AI w sieciach tajnych

The U.S. Department of Defense announced on May 1, 2026 that it has signed agreements with NVIDIA, Microsoft, Amazon Web Services, and Reflection AI to deploy their artificial intelligence technologies and models on its classified military networks. The deals cover Impact Level 6 and Impact Level 7 environments — the highest security classifications in the DOD information infrastructure. The move accelerates the Pentagon diversification strategy following its high-profile legal dispute with Anthropic, which refused to grant the military unrestricted access to its AI models.

Key Takeaways

DOD signed AI deployment agreements with NVIDIA, Microsoft, AWS, and Reflection AI for IL6/IL7 classified networks
Earlier agreements with Google, SpaceX, and OpenAI already existed — Pentagon is deliberately building a multi-vendor AI portfolio
The Anthropic dispute over unrestricted use for autonomous weapons and surveillance was the catalyst for accelerated diversification
More than 1.3 million DOD personnel already use GenAI.mil for unclassified tasks
Pentagon stated goal: establish the U.S. military as an AI-first fighting force

The Anthropic Dispute as Catalyst

The Pentagon had previously contracted with Anthropic for access to its Claude models for military applications. The relationship broke down when the DOD sought unrestricted use — including for autonomous weapons systems and large-scale domestic surveillance. Anthropic refused, citing its own usage policies. The dispute escalated to litigation.

In March 2026, Anthropic won a court injunction blocking the Pentagon attempt to label the company a supply-chain risk — a designation that would have effectively excluded it from government contracts. The case remains active. But the political outcome was already clear: the DOD could not afford single-vendor dependency on any AI provider willing to impose usage restrictions.

The New Deals: IL6, IL7, and What They Mean

The agreements cover IL6 and IL7 environments — the highest tiers of DOD data classification. IL6 covers Secret-level data; IL7 covers Top Secret/SCI. Deploying AI at these levels means large language models and decision-support systems will have access to the most operationally sensitive information the U.S. military holds.

According to the official DOD statement, the AI systems will support data synthesis, situational understanding, and warfighter decision-making. This covers a broad operational range — from intelligence analysis to logistics optimization and mission planning.

NVIDIA contributes compute infrastructure — its GPU hardware and NVIDIA AI Enterprise platform are the default standard for training and running large models in government-grade environments.

Microsoft brings Azure Government, already certified for lower classification levels and now extended upward. AWS GovCloud has operated classified workloads for years — the new agreement extends its role deeper into national security applications. Reflection AI enters this group as a specialist in safety-oriented and reasoning-focused model development.

GenAI.mil and Deployment Scale

The Pentagon disclosed that more than 1.3 million DOD personnel currently use its GenAI.mil platform. The platform provides access to LLMs and AI tools within government-approved cloud environments and is designed primarily for unclassified tasks: research, document drafting, and data analysis.

The new agreements extend this ecosystem into classified networks — a qualitative shift. Previously available tools operated on publicly available or low-classification documents. Access at IL6/IL7 means working with operational intelligence, mission plans, and materials classified as critical to national security.

The DOD statement explicitly frames the strategy around avoiding vendor lock-in: no single company will hold an exclusive position in this market. Each future procurement cycle will remain open to new entrants.

Why This Matters

The Pentagon deals with NVIDIA, Microsoft, AWS, and Reflection AI are not solely a national security story. They signal how mature institutional deployment of large AI models looks at scale — and what conditions vendors must accept to participate.

The scale is significant. 1.3 million users on a single AI platform represents one of the largest known enterprise deployments of generative AI globally. When that environment extends to classified networks, it becomes a precedent for how governments worldwide will structure their AI infrastructure.

The Anthropic dispute established a new category of legal and contractual risk: the possibility that AI labs will impose usage restrictions that conflict with a government client operational requirements. Every major AI company now faces this tension. The multi-vendor architecture the Pentagon is deliberately constructing will influence procurement strategies far beyond the DOD.

What's Next?

The Anthropic vs. Pentagon litigation continues; a ruling could establish precedent for AI model usage terms in government contracts
DOD has signaled further expansion of its AI vendor portfolio — the current group of seven companies is unlikely to be the final list
IL6/IL7 deployments are expected to reach operational status within 6–12 months; effectiveness assessments anticipated in 2027