The parallels of the "attackers" and "defenders" is going to be about how delusional the predictive algorithms they're running.
And reminder: LLMs arn't very good at self-reflective predictions.