IDK, I guess we will just disagree, as I feel like I’m reading a completely separate blog than everyone else.

As the other read makes zero sense. A model cannot be “protected” because it cannot be damaged, it is frozen in place, its state does not change. Anthropic seems to be talking about people “abusing its models” for harmful output, but people here have read it the other way just because of English language ambiguity.