This doesn't clear up the confusion for me. That mostly pushes my question to "what are misalignment related technical problems?" Is the problem of an AI escaping a server and aligning with North Korea a technical or a political problem? How could we tell? Is this still in the regime where we are using AIs as tools, or are you talking about a regime where AIs are autonomous agents?