The most safety-focused AI lab on the planet just got blacklisted by the Pentagon for being too safe. On May 1, the U.S. Department of Defense signed IL6/IL7 classified AI deals with eight companies. Anthropic — the one that publishes its safety policy — was not on the list.

IL6 and IL7 are the Pentagon's classified-data tiers: IL6 covers Secret-level data (operational plans, weapons specs), IL7 covers Top Secret and Special Access Programs. They are the highest commercially-accessible authorization levels in U.S. government cloud, and getting cleared costs years and tens of millions per vendor.

I have been tracking this since the "supply chain risk" label leaked in February, and the May 1 announcement closes the loop in a way that should change how you think about building on frontier AI. The reason Anthropic is locked out is not technical.…