This is a legitimate cause for concern, and it highlights the critical question at the center of our current debates over artificial intelligence and warfare—whether humans could ever train and test intelligent machines enough to trust them to close the kill chain without a person in the loop or directly in control.
