Well, you are not the Pentagon.

I presume they might view their internal data being used to train a private company's training sets as a concern.

I think, assume. With the Military, with the goal of killing people, that there is assumption of more human judgment being involved. Beyond even the issue of a man-in-the-loop for targeting systems. But even generally, that these are big decisions that shouldn't be farmed out completely.

How many of your tasks decide whether entire populations of human beings live or die?