Who's at fault when: Your CloowdBot reads an angry email that you sent about how much you hate Person X and jokingly hope AI takes care of them, only for it to orchestrate such a plan.
How about when your CloowdBot convinces someone else's AI to orchestrate it?
Etc