Curiosity is both the triumph and the curse of mankind. Most technological or scientific advances have been and will in all likelihood always be harnessed for military purposes and there is no reason to believe that artificial intelligence will not. Even so, this does not inevitably herald an impending military ‘robopocalypse’ with legions of killer robots dehumanising the battlefield and defying international law. However, once the Pandora’s Box of military artificial intelligence is opened, it will be too late to address external, internal and ethical concerns arising from military artificial intelligence. Consequently, in order to govern and regulate the development of military artificial autonomy and prevent uncontrollable proliferation of and a potential arms race in autonomous weapons, a constraining-enabling arms control regime needs to be promptly negotiated and legally implemented, striking a functional and legitimate balance between military and non-military interests and concerns.