How do you know it isn’t happening already? World powers have been using AI assisted battle scenario planning for at least a decade already… how would we even know, if some of those AIs decided to appear to optimize for their handler’s goals, but actually aim for their own ones?