The order establishes a framework to vet advanced AI's national security risks up to a month before public release