Envelop our warfighters with advanced AI across all domains
In a move that marks a significant threshold in the militarization of artificial intelligence, the U.S. Department of Defense has certified eight major technology firms — among them Google, OpenAI, Microsoft, and Amazon — to deploy AI systems on its most sensitive classified networks. The decision reflects not a cautious experiment but a deliberate strategic pivot: the Pentagon is seeking to make artificial intelligence foundational to how its forces perceive, decide, and act. It is a moment that asks old questions anew — about the nature of human judgment in warfare, and what it means to surround soldiers not just with tools, but with thinking machines.
- The Pentagon has cleared eight tech giants, including Google, OpenAI, and SpaceX, to run AI on secret and top-secret military networks — a security threshold that few commercial systems have ever crossed.
- GenAI.mil, launched just five months ago on Google's Gemini model, has already drawn 1.3 million military users generating tens of millions of prompts, signaling an adoption curve that has outpaced most institutional technology rollouts.
- Rather than anchoring itself to a single vendor, the DoD is deliberately assembling a multi-provider AI ecosystem — Thunderforge for planning, Grok for operations, Claude Mythos at the NSA — hedging against dependency and technical failure alike.
- With $33.7 billion earmarked for science, technology, and autonomous systems in its 2026 budget request, the Pentagon is marshaling resources at a scale that signals this transformation is structural, not supplemental.
- The administration's own language — promising to 'envelop' warfighters with AI — suggests the goal is not augmentation at the margins but an ambient, foundational reordering of how military decisions are made.
On Friday, the Department of Defense announced it had certified eight technology companies — Google, OpenAI, Microsoft, Amazon Web Services, Nvidia, SpaceX, Oracle, and a smaller firm called Reflection — to run AI systems on its most sensitive classified networks. These networks operate at the highest tiers of the Pentagon's security framework, handling intelligence so restricted that access requires strict clearances, network isolation, and controls that dwarf anything in commercial deployment. Clearing this bar is not routine. That eight major firms have done so at once signals how urgently the Pentagon is moving to embed AI into its operational core.
The department framed the certifications as essential to its strategic vision, describing the goal as transforming the U.S. military into an 'AI-first fighting force' capable of maintaining decision superiority across all domains of warfare. At the center of this effort is GenAI.mil, a platform launched in December built on Google's Gemini model. In just five months, more than 1.3 million military personnel have used it, generating tens of millions of prompts and deploying hundreds of thousands of AI agents. Crucially, the Pentagon built the platform to support multiple AI providers — a deliberate hedge against the risks of single-vendor dependency.
This certification is one thread in a much larger weave. Over the past year, the DoD has contracted with Scale AI to build Thunderforge, a military planning system; integrated xAI's Grok into operations; struck a classified AI agreement with Google; and seen the NSA begin deploying Anthropic's Claude Mythos on restricted networks. The picture is one of deliberate, multi-pronged integration — not a bet on any single technology, but the construction of an AI ecosystem across the force.
The financial commitment behind this transformation is substantial. The Pentagon's 2026 budget request includes $33.7 billion specifically for science, technology, and autonomous systems. Amazon Web Services, one of the eight certified firms, described its commitment to military modernization as a long-term partnership spanning more than a decade. The administration's own framing was pointed: the goal, it said, is to 'envelop' warfighters with advanced AI — not merely to equip them, but to make artificial intelligence ambient and foundational to how they work.
On Friday, the Department of Defense announced it had cleared eight technology companies to run artificial intelligence systems on its most sensitive military networks. The roster included names familiar to anyone tracking the AI industry: Google, OpenAI, Microsoft, Amazon Web Services, Nvidia, SpaceX, Oracle, and a smaller firm called Reflection. The certification opens the door for these companies' AI tools to operate on networks handling classified information at the highest levels—the kind of data that shapes decisions about national security.
The Pentagon's security framework divides classified networks into tiers. Impact Level 6 covers secret-level information; Impact Level 7 handles the most restricted intelligence, the kind designed for only the most sensitive national security systems. For AI to run on these networks, it must operate within tightly controlled infrastructure, with strict access controls, network isolation, and clearance requirements that would make most commercial deployments look like open playgrounds. The fact that eight major tech firms have now cleared this bar signals how seriously the Pentagon is moving to embed AI into its operational core.
The department framed the move as essential to its strategic vision. "These agreements accelerate the transformation toward establishing the United States military as an AI-first fighting force," the Pentagon said in its announcement, adding that the systems would strengthen warfighters' ability to maintain what it called decision superiority across all domains of warfare. The language was direct: this is not about incremental improvement. This is about reshaping how the military thinks and acts.
The centerpiece of this effort is GenAI.mil, a platform the Pentagon launched in December built on Google's Gemini model. In just five months, the numbers have been striking. More than 1.3 million military personnel have used the system. They have generated tens of millions of prompts. They have deployed hundreds of thousands of AI agents across the force. The platform is designed to support data analysis, situational awareness, and decision-making—the cognitive work that sits at the heart of military operations. Notably, the Pentagon built GenAI.mil to work with multiple AI providers rather than locking itself into a single vendor, a hedge against both technical failure and the kind of dependency that comes from relying on one company's technology.
This certification is not happening in isolation. Over the past year, the Pentagon has been steadily expanding its AI partnerships. In March 2025, it contracted with Scale AI to build Thunderforge, a planning system. In July, it signed a deal to integrate xAI's Grok model into military operations. Last month, it reached another agreement with Google for classified AI work. Separately, the National Security Agency has begun deploying Anthropic's Claude Mythos on classified networks, even as a dispute with Anthropic continues in the background. The picture that emerges is one of deliberate, multi-pronged integration—the Pentagon is not betting on any single technology or company, but rather building a ecosystem of AI tools across its operations.
The financial scale of this effort is substantial, though the Pentagon did not disclose the specific value of these new agreements. In its 2026 budget request, the department is seeking $961.6 billion overall, with $33.7 billion earmarked specifically for science and technology and autonomous systems. That figure alone suggests the resources being marshaled for this transformation.
Amazon Web Services, one of the eight certified companies, signaled its commitment to the effort. "For more than a decade, AWS has been committed to supporting our nation's military," a company spokesperson told reporters, adding that the firm looked forward to continuing to support the Pentagon's modernization efforts. The language was measured, but the implication was clear: this is a long-term partnership, and AWS sees itself as central to how the military will operate in the years ahead.
The Pentagon's own framing tied the effort to the current administration's priorities. "As mandated by President Trump and Secretary Hegseth, the Department will continue to envelop our warfighters with advanced AI to meet the unprecedented emerging threats of tomorrow," the department said. The word "envelop" is worth noting—it suggests not just equipping soldiers with AI tools, but surrounding them with AI, making it ambient, foundational to how they work. What happens next will depend on how quickly these systems can be integrated into actual military operations, and whether the promised gains in decision-making and situational awareness materialize at scale.
Notable Quotes
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.— Department of Defense
For more than a decade, AWS has been committed to supporting our nation's military and ensuring that our warfighters and defense partners have access to the best technology at the best value.— Tim Barrett, Amazon Web Services spokesperson
The Hearth Conversation Another angle on the story
Why does the Pentagon need to certify these companies separately? Don't they already work with the military?
They do, but classified networks are a different animal entirely. You can't just plug commercial AI into a system handling top-secret intelligence. The certification process ensures the technology, the infrastructure, and the company's security practices can all handle that level of sensitivity. It's the difference between using Google Maps and building a navigation system for nuclear submarines.
The numbers around GenAI.mil are striking—1.3 million users in five months. Is that adoption or saturation?
It's both. That's a massive portion of the military workforce. But the real question is what they're actually using it for. Tens of millions of prompts could mean anything from routine data queries to mission-critical analysis. The Pentagon isn't saying much about that breakdown.
Why does the Pentagon insist on multiple vendors instead of picking the best one?
Risk distribution. If you lock yourself into one company's technology and something goes wrong—a security breach, a business dispute, a technical failure—you're exposed. By spreading the load across Google, OpenAI, Microsoft, and others, they're hedging. It also keeps any single company from having too much leverage over military operations.
The NSA is deploying Claude on classified networks while there's a dispute with Anthropic. That seems contradictory.
It does. It suggests the technology is valuable enough that the dispute hasn't stopped deployment. But it also hints at tension beneath the surface—the military needs these tools, but the relationships with the companies building them are more complicated than the official announcements suggest.
What's the real risk here?
Dependency. Once you've trained a million soldiers to use AI for decision-making, you've created a system that can't easily be unwound. If there's a security failure, or if a company decides to change its terms, the military is stuck. You've also created a massive surface area for adversaries to target.