i guess adverserial attacks towards the employed ai-model or some form of physical obstacle (net) would be most efficient but emp or directed energy weapons are

i guess adverserial attacks towards the employed ai-model or some form of physical obstacle (net) would be most efficient but emp or directed energy weapons are probably more reliable