You can see that the use cases above (commentary, criticism, news reporting and scholarly reports) does not qualify LLM companies to use or train their models

Seems quite obvious that the text you quoted refers exclusively to plagiarism. This does not include things like being inspired by it, referencing it, parodying it and of course not training AI either, because what matters is whether the result is protected content.

You can argue that memorizing and sharing training data is a copyright violation, and that's a fair point, but it's also worth noting that this is very much a minority, accidental and is being addressed.