
The Pentagon Asked OpenAI for an AI That Rarely Says No | Sep 11, 2026
Leaked contract records reveal the Pentagon wanted a version of OpenAI's models built to refuse military commands as rarely as possible — and the same documents show why Anthropic didn't make the cut.
This week, Marcus and Sam dig into FOIA-obtained documents from The Intercept showing the Department of Defense asked OpenAI for AI with "minimal refusal rates" against military tasking, while Anthropic was formally designated a "supply chain risk" by Defense Secretary Pete Hegseth after refusing to drop its restrictions on autonomous weapons and domestic surveillance. OpenAI, Google, and xAI…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.