What makes you think you can measure the % say an AI has in the decision to kill a human? Even if we pretend "it had 100% say" was true it wouldnt matter, it would still be a human that ordered the deployment and be responsible for the decision.