The perverse reversal here, is that if Anthropic had not turned down working on MILSPEC data they would probably have been told to welcome the engagement so the state could functionally understand what foreign strategic interests wanted in AI.