It’s not their IP, they never created data they were training on in the first place, at best they derived existing data without asking if they can commercialized it