That seems sensible.
Even a hypothetically true artificial general intelligence would still not be a moral agent, thus it cannot be held responsible for its actions; as such, whoever deploys and maintains it should be held responsible. That’s doubly true with LLMs as they aren’t even intelligent to begin with.
Pretty much. Once speakers start using the word, and expecting others to understand it, it’s already part of the lexicon of that language. Specially if you see signs of phonetic adaptation, like /ø/ becoming /u:/ in a language with no /ø/ (see: “lieu”) - and yet it’s exactly why people complain about those words.
And this sort of complain isn’t even new. Nor the backslash agianst it, as Catullus 84 shows for Latin and Greek.