My compliance brain says no, but the fact that models get trained with data they obtain without explicit permission makes says that finders keepers would be the relevant case law.