Well, to a point that's true: there is a not-clearly-defined line about what's copyrightable when it comes to the amount of text that's reproduced without explicit permission (as a matter of fact, that might be the same thing OpenAI is using as a loophole to consider their model immune to copyright claims).

Basically, I can't copyright an "article" which only says "This is an article." — there is high probability of someone coming up with the exact same wording, so you can't get protections for it. How much is too much is definitely left up to the interpretation.

Finally, unless one of those contributors takes him to court (external party like EFF or FSF can't protect your rights, you have to do it yourself), it's unlikely he should care. If one or two people do, there's also a relatively cheap option of rewriting just those parts too: court won't try to challenge other possible license issues.

he claims it's 99% by him and paid contributors. to verify you'd have to ask any major contributors if they have been paid.