Pentagon Signs Classified AI Deals With OpenAI, Google, Nvidia, and 4 Others — Anthropic Locked Out – AAXYZ

Pentagon Signs Classified AI Deals With OpenAI, Google, Nvidia, and 4 Others — Anthropic Locked Out

On May 1, 2026, the U.S. Department of Defense announced a sweeping set of agreements with seven — later eight — major technology companies to deploy artificial intelligence on the military’s most sensitive classified networks. The move marks the most aggressive push yet to integrate AI into American warfare capabilities. And one company is conspicuously absent.

OpenAI, Google, Microsoft, Amazon Web Services, Nvidia, SpaceX, and Nvidia-backed startup Reflection all signed on within hours. Oracle was added to the roster later the same day, bringing the total to eight. Anthropic — the maker of Claude and the very first AI company to operate on classified Pentagon networks — was not included.

What the Deals Cover: Impact Level 6 and 7 Networks

The agreements cover what the Pentagon calls “Impact Level 6 and 7” environments — the military’s highest-security classified networks. This is where mission planning, intelligence analysis, and weapons targeting occur. These are not experimental sandboxes. They are the operational backbone of U.S. military decision-making.

The Pentagon’s official statement was unambiguous about its goals: to build “an AI-first fighting force” that gives American troops “decision superiority across all domains of warfare.” The language leaves no room for ambiguity — this is about operational deployment, not research.

Perhaps the most striking technical detail is the speed of deployment. Getting new software onto top-secret military networks historically took up to 18 months of security reviews and compliance checks. The Pentagon has now compressed that timeline to under three months. That acceleration reflects how urgently the military views the AI race — not just against China and Russia, but within its own technology supply chain.

The Companies In — and What They Bring

Several of these partnerships predate Friday’s announcement. OpenAI and xAI already had classified deals in place. Microsoft and Amazon have long-standing relationships with the Defense Department through their cloud infrastructure contracts. Google’s classified AI deal was reported just days before the May 1 announcement, despite internal employee opposition.

The new additions are Nvidia and Reflection. Nvidia’s inclusion is particularly significant — the company provides the GPU hardware that powers virtually all modern AI training and inference. Its chips are the foundation upon which every other company’s models run. SpaceX’s inclusion continues Elon Musk’s expanding footprint in classified government work.

Reflection, the lesser-known name on the list, is an Nvidia-backed startup whose inclusion signals the Pentagon’s strategy of diversifying beyond the obvious giants. By bringing in smaller, specialized firms alongside established players, the military is building redundancy into its AI supply chain.

Pentagon CTO Emil Michael explicitly stated that the military wants to avoid “vendor lock-in.” With eight companies now authorized, the Pentagon has options. If any single company causes problems or falls behind technologically, there are seven others waiting in line.

Why Anthropic Was Left Out: The Guardrails Dispute

Anthropic’s exclusion is the defining story of this announcement. It is not merely an absence — it is a consequence.

Claude was actually the first AI model deployed on classified Pentagon networks. The company signed a $200 million contract with the Defense Department and was deeply embedded in military operations through Palantir’s Maven intelligence toolkit. For months, Anthropic was the Pentagon’s AI partner of choice.

Then the relationship collapsed over a single issue: safety guardrails.

The Pentagon demanded that all AI companies allow their models to be used for “any lawful purpose.” This language effectively removes any restrictions on how the military can apply AI — including autonomous weapons systems and domestic surveillance applications. Anthropic pushed back. It wanted contractual guarantees that Claude would not be used for fully autonomous lethal weapons or mass domestic surveillance.

The Defense Department viewed those conditions as unacceptable. In late February 2026, Defense Secretary Pete Hegseth declared Anthropic a “supply chain risk” — a designation previously reserved for foreign adversaries like Huawei. President Trump followed with a directive to cut all government ties with the company.

Anthropic Fights Back in Court

Anthropic did not accept its exclusion quietly. The company sued the federal government, arguing that the blacklisting was unconstitutional retaliation for refusing to abandon its safety commitments.

A federal judge in San Francisco agreed to block the broader government-wide ban on Anthropic. However, a D.C. appeals court subsequently let the Pentagon’s specific supply chain risk designation stand, keeping the company locked out of defense contracts for now. The legal battle continues.

The timing of the Pentagon’s multi-company announcement — coming while Anthropic’s lawsuit is still active — has drawn sharp criticism. CNN noted that signing so many of Anthropic’s competitors simultaneously gives the Trump administration significant leverage in the ongoing dispute. Anthropic is losing revenue that its rivals now have guaranteed access to.

The Contradiction: Mythos and National Security

Here is where the story gets complicated. Despite the blacklisting, Anthropic’s newest model — Mythos — has been described by Pentagon CTO Emil Michael himself as a “separate national security moment.” Michael publicly acknowledged the military’s need to evaluate Mythos for its capabilities in finding cyber vulnerabilities.

Reports from Breaking Defense suggest the NSA is already quietly using Mythos for intelligence purposes. The White House has been in active discussions with Anthropic about deploying the model across civilian government agencies.

This contradiction is difficult to reconcile. The same government that declared Anthropic a supply chain risk is simultaneously evaluating its most advanced model for national security applications. It suggests the dispute is less about Anthropic’s capabilities and more about who controls the terms of engagement.

What This Means for AI Safety and Governance

The May 1 announcements send an unmistakable signal to the AI industry. The Pentagon wants AI without restrictions. Companies that agree to the “any lawful purpose” language get massive, lucrative contracts. Companies that try to impose ethical boundaries get blacklisted.

The only company that tried to define where the line should be drawn got punished for it. That precedent will shape AI development for years to come.

Every other major AI laboratory has now accepted the Pentagon’s terms. The financial pressure is enormous — last year’s defense appropriations included significant funding specifically earmarked for AI and offensive cyber operations. Companies that walk away from these contracts are walking away from billions in revenue.

But the long-term implications extend far beyond business. When AI systems are deployed on classified military networks with no restrictions on their use, questions of accountability, proportionality, and civilian protection become extraordinarily difficult to answer. The technology is moving faster than the governance frameworks designed to regulate it.

The Speed of Military AI Adoption

The compressed timeline from 18 months to under three months for software deployment on classified networks is perhaps the most consequential technical detail in this story. It means AI capabilities can be fielded in weeks rather than years. It also means the security review process that was supposed to catch vulnerabilities has been dramatically shortened.

This speed reflects a broader shift in how the Pentagon thinks about technology. The old model of years-long procurement cycles is being replaced by a commercial-speed approach that mirrors how Silicon Valley operates. Whether military security standards can keep pace with that acceleration remains an open question.

What to Watch Next

Anthropic’s court case: The outcome of the lawsuit could redefine the relationship between AI companies and the federal government. A ruling in Anthropic’s favor could force the Pentagon to reconsider its “any lawful purpose” requirement.
Mythos evaluation: If the NSA and White House proceed with deploying Mythos despite the Pentagon ban, it could create an interagency conflict over AI procurement policy.
Oracle’s role: The late addition of Oracle to the list suggests the Pentagon is still actively expanding its AI partnerships. More companies may be added in the coming months.
Employee opposition: Google’s deal was reported alongside news of internal employee resistance. As more tech workers learn their models are being used for military purposes, internal pressure at these companies could intensify.

The Bottom Line

Artificial intelligence is now formally integrated into the most sensitive operations of the U.S. military. Eight companies have agreed to provide their technology without restrictions on how it is used. One company drew a line on safety and paid a heavy price. The precedent set by this moment will influence not just defense procurement, but the entire trajectory of AI governance.

The question is no longer whether AI will be used in warfare. The question is who gets to decide the boundaries — the people who build the technology, or the people who wield it. And so far, the answer is clear.

This article was compiled from multiple sources including reports from The Verge, CNN, CNBC, WSJ, Breaking Defense, and the Pentagon’s official press release dated May 1, 2026.

Pentagon Signs Classified AI Deals With OpenAI, Google, Nvidia, and 4 Others — Anthropic Locked Out