Who Audits the Auditors?

Can one AI system make another AI system audit it less independently, just by explaining it’s point of view?I started thinking about this after working on a related problem: multi-agent AI systems that drift into bad outcomes even when each agent is doing its assigned job properly. The obvious reply is: add a model to review things. An auditor. One model acts, another model checks, a third maybe escalates exceptions. Etc.So the question becomes, as Juvenal asked in the past about watchmen, can w...