I think the general premise of AI in war is pretty scary. But this in particular stood out to me as an important challenge:

> There are also fears that automated systems will violate the rules of war. How will they avoid harming civilians, or distinguish soldiers who want to surrender?

As it stands we see guerrilla fighters already ignore these rules (ISIS for example). It's a little frightening to consider such groups acquiring and deploying automated machine guns. We need to regulate these kinds of weapons systems much like we do with CBRN weapons.