Pentagon AI Deals: U.S. Awards Classified Contracts to Seven Tech Giants, Anthropic Left Out

Pentagon Awards AI Contracts to Seven Tech Giants for Classified Networks

On May 1, 2026, the U.S. Department of Defense officially announced contracts with seven technology companies to use artificial intelligence in its classified networks. The list includes OpenAI, Google, Microsoft, Amazon Web Services, Nvidia, Elon Musk's SpaceX, and the young AI startup Reflection. The aim is to embed language models, agent software, and image and data analysis directly into military systems.

Anthropic Excluded: Dispute Over Safety Guardrails for Autonomous Weapons

What stands out is who is missing: Anthropic, the maker of the Claude models. According to multiple U.S. outlets, the Trump administration effectively blacklisted the company after Anthropic insisted that the Pentagon adopt specific safety guardrails. CNN and the Washington Post report that the dispute centered on autonomous weapons and mass surveillance. Anthropic had a 200-million-dollar Pentagon contract awarded in 2025 that the administration severed. The company has filed lawsuits in San Francisco and Washington, D.C., which are still ongoing.

Three Structural Shifts in the U.S. AI Market

The deals signal three shifts. First, generative AI is moving into the classified domain, meaning intelligence and military networks. Second, the U.S. is deliberately spreading work across many vendors rather than crowning a single champion. Third, terms of use are becoming a competitive lever: vendors that push hard on safety clauses now risk losing contracts.

Why it matters

For organizations outside defense, this matters because the same vendors serve civilian markets. The tooling, audit logs, and identity controls built for the Pentagon program will likely make their way into cloud offerings for energy, healthcare, and financial services. The Anthropic episode also shows that ethical commitments can have a price tag: vendors that draw hard lines lose concrete contracts to competitors that do not.

Still open is exactly how the Pentagon will harden these models. According to public reporting, the contracts foresee separated enclaves, vetted training data, and logged access. Detailed evaluation procedures are not public for security reasons.

Practical example

A German utility runs Microsoft Azure and Google Cloud in its control center. If Microsoft and Google now build additional security features for the Pentagon, such as signed model containers, stricter identity checks, and complete audit trails, those features will land in standard tiers over the coming quarters. The utility will not have to build them itself, but can benefit once it connects AI agents to its SCADA systems. Companies betting on Anthropic today should plan backup options in case regulatory bans spill into other markets.