DefenseTech talks about a new legal interpretation of the rules governing unmanned combat. Rules are currently interpreted as meaning that an unmanned robotic weapon system cannot fire without a human operator in the control loop. A new 'interpretation', or loop hole, is to re-program robots to aim only at weapons. They should be allowed to fire at weapons under the law. While a novel interpretation, I think it would be relatively easy to conceal the shape of a weapon if you knew that a robot would only fire on a weapon.
The idea is the brain child of John S. Canning, chief engineer at the Naval Surface Warfare Center.