I'm assuming these won't be powered by AI, that will come later. The first generation will probably be controlled by humans.

But I've played video games where the AI of the security guards completely lost it's mind and started slaughtering all the unarmed NPC civilians in an area, when I, the player, was the only one who was being naughty. Basically the security guards in the game existed only to attack the player if they did something bad like stealing things or breaking into restricted areas of the map or attacking innocent characters, all of which I was doing. (immersive sim games are awesome.) But, instead of attacking me, the guards wiped out all the unarmed NPCs in the area. Crazy glitch! Especially because the unarmed characters couldn't have attacked anyone. I think what must have happened in the game, is that one of the security guards found one of the unconscious NPCs I knocked out and somehow attributed it to the civilian team, and all hell broke loose.

Ideally, we would not be stupid enough to potentially allow scenarios (armed and dangerous, buggy/rogue AI) like the above to play out in the real world.