Pentagon signs classified-network AI agreements with major vendors, excluding Anthropic

What happened

The Defense Department announced classified-network AI agreements on May 1, 2026. Public reporting has differed on the vendor count. Reuters and AP listed seven providers, while the department-linked release and some defense coverage list eight, including Oracle. Named vendors across the coverage include SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, AWS, and Oracle.

The department framed the agreements as part of a push to make the U.S. military an AI-first force across all domains of warfare.

Anthropic was excluded by name. The dispute reportedly centers on Anthropic's restrictions around certain military, surveillance, and autonomous-weapons uses of Claude. The Pentagon labeled Anthropic a supply chain risk, a label rarely applied outside foreign adversary cases. Anthropic is suing the department over the move. The White House has reopened talks with the company in recent weeks.

Why it matters

The agreements split the AI industry into two camps. Vendors that accept broad military use can compete for classified-network work. Vendors with stricter use restrictions may face limits on certain classified-network opportunities.

For the wider market, the agreements give the Defense Department multiple AI suppliers at once. A senior official said the department will never again rely on a single AI provider.

MintedBrain take

For users, this is mostly a story about AI policy, not products. But it shows how vendor choice now carries political and ethical weight. Companies that buy AI need to think about which provider they align with, not just which model performs best on benchmarks.