lukeprog comments on My Overview of the AI Alignment Landscape: Threat Models