Just like every other major model out there.

Wow From the article wrote a script to remove copyright info, including the word “copyright” and “acknowledgments,” from ebooks in LibGen.

suggests that Meta strips [copyright information] ......

.....it torrented LibGen,

Incoming: Another round of "OMG that's so illegal" from the same people that say copyright shouldn't exist.

Setting aside the question of who holds it, is it an inconsistent position? If a broadening of fair use is necessary for developing large language models, it's perfectly consistent to argue it should also extend to artists/archivists/derivative works in general, otherwise you're staking out the position that pirating one book is wrong but downloading Books3 is fine

It's all so tiresome.