The Pentagon's announcement confirmed agreements with eight major technology organisations, cleared to deploy their AI capabilities on the Department of Defense's classified networks for what the DoD described as "lawful operational use." The initial announcement covered seven firms — Amazon Web Services, Google, Microsoft, OpenAI, SpaceX, Nvidia, and NVIDIA-backed startup Reflection AI — with Oracle added hours later in a follow-up statement by the Pentagon CTO's office.

The DoD's official statement set out the ambition in unambiguous terms: "These agreements accelerate the transformation toward establishing the United States military as an AI-first fighting force and will strengthen our warfighters' ability to maintain decision superiority across all domains of warfare." The Pentagon also committed to building an architecture that prevents AI vendor lock-in, explicitly stating its intent to maintain long-term flexibility across a resilient and diverse American technology stack.

These deals do not represent the DoD's first steps into AI — the Pentagon already held existing AI contracts with companies including Palantir and OpenAI prior to these agreements. But the breadth and speed of this latest wave of clearances — spanning hyperscalers, chip manufacturers, aerospace companies, and AI-native startups simultaneously — represents a step-change in scale and strategic intent.

The timing and breadth of these agreements cannot be separated from the Pentagon's highly publicised and ongoing legal dispute with AI safety company Anthropic. The DoD sought unrestricted use of Anthropic's AI models on its classified networks — but Anthropic refused, insisting on contractual guardrails that would prevent its technology from being used for domestic mass surveillance and fully autonomous weapons systems. The Pentagon's response was to classify Anthropic as a "supply-chain risk" — a move Anthropic challenged in court and, in March, successfully overturned, winning an injunction against the designation.

The dispute with Anthropic directly accelerated the DoD's diversification strategy. Rather than depending on any single AI provider, the Pentagon moved rapidly to assemble a broad coalition of AI partners — each offering different capabilities and, crucially, each willing to operate under the DoD's terms without imposing usage restrictions. The eight companies now cleared represent precisely this architecture of flexibility: cloud infrastructure, frontier AI models, chip compute, and space-based connectivity — all from separate providers, all deployable on classified networks simultaneously.

Notably, Google expanded the Pentagon's access to its AI specifically in the wake of Anthropic's refusal — a stark illustration of how the Anthropic dispute reshaped the competitive dynamics of the entire US defence AI market almost overnight.

The Anthropic dispute crystallised the central ethical tension in military AI: who sets the limits on how AI is used in warfare, and who enforces them? Anthropic's position — that an AI lab has the right and responsibility to impose guardrails on how its technology is used, even by sovereign governments — is a genuinely novel assertion in the history of defence contracting. No weapons manufacturer has historically been able to dictate terms of use to the US military. That an AI company attempted to — and won a court injunction in the process — marks a significant moment in the evolving relationship between technology firms and state power.

The eight companies that did agree to DoD terms have chosen a different path. None has publicly disclosed the specific usage restrictions — if any — agreed upon in their individual deals. The lack of transparency around the precise terms of these agreements is itself a point of concern for civil society organisations and researchers tracking the responsible development of military AI. As AI systems become more capable and more deeply embedded in military decision-making, the governance frameworks that constrain their use will become one of the most consequential policy questions of the coming decade.

Key Takeaways

• →The US Department of Defense has signed AI deployment agreements with eight major technology companies — AWS, Google, Microsoft, Nvidia, OpenAI, SpaceX, Oracle, and Reflection AI — for use on classified military networks
• →The Pentagon's stated goal is to establish the US military as an AI-first fighting force — with decision superiority across air, land, sea, space, and cyber domains
• →The agreements were accelerated by the Pentagon's high-profile dispute with Anthropic, which refused to grant unrestricted use of its AI models for military purposes and won a court injunction against being labelled a "supply-chain risk"
• →The DoD is deliberately building a multi-vendor AI architecture to avoid lock-in — spanning cloud infrastructure, frontier models, GPU compute, satellite connectivity, and AI-native startups
• →Google expanded Pentagon AI access in the direct wake of Anthropic's refusal — and simultaneously exited a $100 million drone swarm programme, illustrating the nuanced and case-by-case nature of tech-defence relationships
• →The deals raise fundamental questions about AI governance in warfare — who sets the ethical limits, who enforces them, and how transparent the terms of military AI agreements should be in a democratic society