> That the majority of comments under this article are looking at red-teaming as a containment protocol for rogue AGI (even though the actual text basically nev

> That the majority of comments under this article are looking at red-teaming as a containment protocol for rogue AGI (even though the actual text basically never talks about AGI at all) is a disturbing reminder of how uneducated developers are about boring, practical everyday risks that are present in everyday current models. Yes, but it might also be a reminder of how many people comment on HN without reading an article :)