Who specifies the boundaries within which artificial intelligence operates? For instance, would you want an AI commander to have access to battlefield nukes? What constraints would you place on their use? How would you specify those constraints in a manner which cannot be subverted by the AI’s ability to shape its understanding?
You have much more faith in the programmers of the computer than I do. Even AI has a layer of human-engineered code underneath it all.
Those are the programmers and army commanders to sort it out