Comment by MarsIronPI

Someone should write a version of this that uses AI to detect whether the command that the AI wants to run is dangerous. Certainly that seems like the current trend in software "engineering".