If someone made a table of one hundred potential targets and rolled ten ten sided dice to pick a row to target, do we blame the decision on the dice, the table, Wizards of the Coast, or the person who decided this was a good way to choose targets?

You’re conflating who chooses the decision maker with the decision making device.

So AI’s output is not so different than a roll of dice? Definitely the people who decided to put in such an incapable and unreliable tool should be held accountable. No dispute in that.