The main problem with the output of LLMs is people not checking results, but there are ethical problems with the effective use of it too. Especially if you're you're dealing with the major platforms, who have shown no qualms allying themselves with fascist governments, surveillance capitalists, or just generally being cavalier about workers rights. (And obviously extracting value from the commons and renting it out on a scale never before seen, but that's the obvious one.)

I would say the same thing about hosting services on AWS. The corporate world does it because it just makes financial sense, but all they care about is outputs. I expect better from individuals who are trying to send a message.