Pentagon awards classified-network AI contracts to seven companies, excludes Anthropic

Summary

The Pentagon announced agreements with seven companies (AWS, Google, Microsoft, Nvidia, OpenAI, SpaceX, Reflection AI) to deploy AI on classified networks (Impact Level 6 and 7). Oracle was later added as the eighth. Anthropic was formally excluded under a supply chain risk designation formalized by Defense Secretary Pete Hegseth in March. The exclusion stems from Anthropic’s refusal to permit Pentagon use of Claude for “all lawful purposes” — language the company argued could enable domestic mass surveillance or fully autonomous weapons. Pentagon CTO Emil Michael told CNBC that Anthropic is still blacklisted but that Mythos is a “separate national security moment.”

Implications

• Mythos / Glasswing thread: Formalizes the institutional split. White House, courts, CISA, and intel community favor Anthropic access; Pentagon CTO opposes. The negotiating and blocking branches now have concrete policy outcomes — contracts awarded to rivals while the ban is challenged in court.
• Enterprise deployment thread: Government classified networks are the highest-stakes enterprise deployment. Seven vendors (plus Oracle) now have deployment experience Anthropic doesn’t. If national security AI becomes the primary enterprise vertical, the exclusion costs more over time.
• Token economics thread: Anthropic’s Q1 2026 revenue share (31.4%, ahead of OpenAI) shows the commercial position isn’t threatened by losing military contracts. The values position is commercially sustainable at current trajectory.
• Reflection AI (NVIDIA-backed, open-source model positioning) is a new entrant worth tracking — first government endorsement of open-source/open-weight models for classified use.