If this exploit has nothing to do with AI, why haven't we heard about it succeeding before? I find it hard to believe it's never been tried.