They're almost definitely trained using an archive, likely taken before they announced the whole API thing. It would be weird if they didn't have backups going back a year.
They're almost definitely trained using an archive, likely taken before they announced the whole API thing. It would be weird if they didn't have backups going back a year.