Pentagon Signs Seven Tech Juggernauts to Build Their “AI-First Fighting Force”

The Pentagon has finalized contracts with seven of the world’s most powerful artificial intelligence companies to embed their systems inside the military’s most sensitive classified networks, the Department of War announced Friday. OpenAI, Google, Microsoft, Amazon Web Services, Nvidia, SpaceX, and the startup Reflection AI will all be permitted to operate inside what the Pentagon designates Impact Level 6 and Impact Level 7 environments — the digital sanctum where war plans are drawn and adversaries are tracked.

The phrase military leaders are using to describe the project is striking. Emil Michael, undersecretary of defense for research and engineering, said the goal is to equip American forces with “a suite of AI tools to maintain an unfair advantage and achieve absolute decision superiority.”

The Pentagon’s official language goes further still, framing the deals as a step toward becoming an “AI-first fighting force.” Whatever one thinks of that doctrine, the country has now committed to it. The contracts moved past the negotiation stage and into operational deployment in a single afternoon.

What makes the story worth reading carefully is not the list of corporate logos. It is the quiet ideological war being fought underneath the announcement — a war over who decides how American military power is used, and whether the people who build the tools get to dictate the terms to those who carry the rifles.

The Anthropic Cautionary Tale

For most of last year, Anthropic’s Claude models held a privileged position inside the Pentagon’s classified environments, supplied through Palantir’s Maven targeting platform. The company’s models were used during the Iran war and the operation that captured Nicolás Maduro. Then Anthropic insisted on contractual language carving out its technology from mass surveillance work and lethal autonomous weapons. The Pentagon’s response was swift and unambiguous. The Defense Department designated Anthropic a supply-chain risk and severed the relationship.

Testifying on Capitol Hill on Thursday, Secretary of War Pete Hegseth called Anthropic CEO Dario Amodei an “ideological lunatic.” It was a calculated insult, and the message was unmistakable. The Pentagon will not allow private AI labs in San Francisco — staffed largely by progressive engineers with strong personal opinions about American foreign policy — to write their own moral side conditions into contracts with the United States military. The ones who attempt it get blacklisted. The ones who comply get billions.

Anthropic is now fighting the designation through two separate legal proceedings. Whether the company prevails in court is almost beside the point. The signal has been received throughout the industry.

Google’s Internal Revolt Collapses

Nowhere has that signal landed more loudly than at Google. In late April, more than 600 Google employees, including senior researchers from the elite DeepMind lab and over twenty directors and vice presidents, signed a letter to CEO Sundar Pichai begging him to refuse any classified Pentagon work. Their argument was that on air-gapped classified networks, Google would have no way to monitor how its Gemini models were being used, and therefore no way to enforce its own ethical commitments.

“We want to see AI benefit humanity; not to see it being used in inhumane or extremely harmful ways.”

The organizers framed it as a continuation of the 2018 Project Maven uprising, when 4,000 Google signatures and a handful of high-profile resignations forced the company to walk away from a Pentagon drone-imagery contract. “Maven is not over,” the organizers declared. “Workers are going to continue organizing against the weaponization of Google’s AI technology until the company draws clear, enforceable lines.”

This time, Google’s leadership did not blink. Pichai signed the deal. The 600 signatures, the DeepMind researchers, the chief scientist publicly invoking the Fourth Amendment on social media — all of it lost to a contract worth a fortune in classified military work. Eight years ago, the activist class inside Big Tech could veto American defense projects with a petition. In 2026, they cannot.

Several things changed between Maven and now. The contracts are larger by orders of magnitude. The Pentagon has demonstrated it will retaliate against refusal by designating dissenters as supply-chain risks. China’s open-source AI labs have made the national security stakes impossible to ignore. And Google itself quietly removed weapons-related language from its AI principles last year, clearing the runway for exactly this moment. The activist employees were the last people in the building who hadn’t gotten the memo.

An “AI-First Fighting Force”

The scale of what the Pentagon is doing deserves a moment of reflection. More than 1.3 million Defense Department personnel have already used GenAI.mil, the military’s official AI platform, generating tens of millions of prompts and deploying hundreds of thousands of agents in just five months. With Friday’s announcement, the most powerful frontier models in existence — closed and open, foundation and applied — will now run inside the networks where targeting decisions, intelligence analysis, and operational planning happen.

The Pentagon’s stated rationale is straightforward. “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 Department of War said in its statement. The multi-vendor approach is designed to prevent dependence on any single company and to keep the architecture flexible as the technology evolves.

Nvidia’s inclusion is especially notable. The chip giant’s contract covers its Nemotron open-source models, designed to let AI agents execute tasks autonomously. CEO Jensen Huang has argued openly that transparent, open-source systems are actually more secure for national security applications because their workings can be inspected and adapted. Administration officials have pointed out that the leading Chinese AI labs are building open models specifically to spread their influence abroad. Without an American open-source alternative, the global standard for military and dual-use AI is being written in Beijing.

Who Controls the Sword

The deeper question this story raises is one that has been simmering in American life for nearly a decade. Who is in charge of the country — the elected government accountable to voters, or the Silicon Valley managerial class accountable to no one in particular?

For years, the answer was ambiguous. Tech companies functioned as a soft fourth branch of government, deciding what speech could be spoken on their platforms, which presidential candidates could communicate with the public, and which Pentagon contracts were morally acceptable for American soldiers to benefit from. The Project Maven climbdown in 2018 was a high-water mark of that arrangement. A coalition of Google engineers, in effect, vetoed an American military program.

The Trump administration has spent the past year demolishing that arrangement, methodically and with intention. The Anthropic blacklisting, the Hegseth testimony, and now the Google employees’ failed petition all point in the same direction. The federal government has reasserted its authority to decide what the American military will use, and the companies that wish to do business with the country’s largest customer will sign on the dotted line and keep their political opinions in their personal time.

This is, properly understood, the restoration of a constitutional norm. The framers vested the war power in the people’s representatives, not in a guild of credentialed engineers. “Where no counsel is, the people fall: but in the multitude of counsellors there is safety,” reads the Hebrew wisdom in Solomon’s collection of sayings — but the counsellors envisioned were elected officials and military professionals, not Bay Area researchers running anonymous petition drives over their lunch hour.

The Risks Worth Naming

None of this is to suggest the AI-first fighting force is without genuine concerns. The Google employees who signed the letter were not all activists. Some were technical experts who understand precisely how these systems can fail, hallucinate, and centralize decisions in ways that humans struggle to audit after the fact. The Pentagon’s “all lawful uses” formulation is a sweeping concession of authority, and the Constitution’s protections against domestic surveillance do not enforce themselves.

Conservatives who have spent years warning about the surveillance state under Democratic administrations have every reason to insist on the same skepticism now. The tools being deployed are unprecedented in their reach. The same systems that can identify a Houthi missile launcher in the Red Sea can, with a few prompt changes, identify an American protester at a courthouse. The technology is morally neutral. The political culture that operates it is not.

The right answer is not to outsource these decisions to Sundar Pichai’s HR department. It is to have them made by elected officials, in the open, with congressional oversight that actually oversees and a press corps that actually questions. The Book of Romans reminds Christians that civil authority “beareth not the sword in vain” — but it bears the sword on behalf of the people, not on behalf of itself.

The Pentagon now holds the sword that Silicon Valley spent a decade building. What comes next will depend on whether the public continues to insist that the people who carry it are actually the ones we elected.