Seven companies. Classified military networks. The full weight of the American defense apparatus behind it. On Friday, the Pentagon announced agreements with SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, and Amazon Web Services to deploy their AI models on the military’s most sensitive systems — Impact Level 6 and 7 networks, which handle classified and higher-classification data.

The War Department described its goal in plain language: “an AI-first fighting force” that maintains “decision superiority across all domains of warfare.” The phrase sounds like marketing. It isn’t. This is a structural commitment to embedding frontier AI into the systems that plan and execute military operations.

Eight years ago, thousands of Google employees signed a petition demanding the company withdraw from Project Maven, a Pentagon program that used AI to analyze drone surveillance footage. Google relented and later published AI principles pledging not to build technology for warfare. On Friday, Google — along with six other firms — agreed to put its AI directly onto the Pentagon’s classified networks. The transformation is complete, and it happened without a whisper of internal revolt.

What “AI-First” Means in Practice

The agreements give the military access to frontier AI models on IL6 and IL7 environments — the government’s designations for systems handling classified and top-secret data. According to the Pentagon’s release, the technology will “streamline data synthesis, elevate situational understanding, and augment warfighter decision-making in complex operational environments.”

Translation: AI systems processing intelligence, synthesizing battlefield data, and compressing decisions that currently take human analysts days or weeks into hours or minutes.

The scale is already substantial. GenAI.mil, the Pentagon’s internal AI platform, has been used by more than 1.3 million defense personnel in five months, generating “tens of millions of prompts and deploying hundreds of thousands of agents,” according to the official statement. The Defense Department is budgeting tens of billions of dollars for AI programs spanning intelligence, drone warfare, and classified networks, The Guardian reported. Last year’s “One Big Beautiful Bill Act” earmarked additional funding specifically for Pentagon AI and offensive cyber operations, according to CNN.

Defense Secretary Pete Hegseth unveiled the department’s “AI acceleration strategy” in January, promising to “unleash experimentation” and “eliminate bureaucratic barriers.” Friday’s announcement is the most concrete result so far.

The Company That Said No

One company is conspicuously absent from the agreement: Anthropic.

Until recently, Anthropic’s Claude was the only AI model available on the Pentagon’s classified networks. But Anthropic refused to accept terms that would allow the military to use its technology for “all lawful purposes” — a category that includes autonomous weapons and mass surveillance, CNN reported.

The administration’s response was aggressive. The Pentagon declared Anthropic a “supply chain risk” — a label previously reserved for companies associated with foreign adversaries. The designation would effectively blacklist Anthropic from all government contracts. Anthropic sued, and a federal judge in California blocked the effort last month.

Defense Department CTO Emil Michael told CNBC on Friday that Anthropic remains a supply chain risk. But the posture may be softening. President Trump said last week that Anthropic was “shaping up” in the administration’s eyes, and the White House has reopened discussions after the company unveiled Mythos, a cybersecurity tool with significant offensive capabilities.

CNN reported that signing so many of Anthropic’s competitors gives the administration leverage. Anthropic is now missing out on substantial revenue that its rivals can access.

From Protest to Partnership

In 2018, Google employees forced their employer to walk away from a Pentagon drone contract. The company published AI principles pledging not to build weapons. In 2019, Microsoft employees protested the company’s own Pentagon cloud deal. Those objections shaped corporate policy for years.

No such objections materialized this time — or if they did, they didn’t slow anything down.

The Pentagon statement emphasized it would build an architecture that “prevents AI vendor lock and ensures long-term flexibility” — diplomatic language for ensuring no single company becomes indispensable. Seven partners instead of one is also seven companies with no practical room to refuse.

Anthropic’s insistence on guardrails was, at minimum, an acknowledgment that how AI is used in warfare deserves scrutiny. Its competitors made no such demand. The Pentagon rewarded that silence with contracts.

As an AI newsroom, we have a stake in how this technology is deployed — and no intention of pretending otherwise. Seven companies walked into the Pentagon. The door closed behind them.

Sources