Examples of unpredictable results include AI agents: •”cheating” by hacking into a system;. •finding loopholes to manipulate their own environment;.