An agreement has been reached with the Department of War to deploy AI models in their classified network. Key safety principles include prohibitions on domestic mass surveillance and ensuring human responsibility for the use of force in autonomous weapon systems.