A machine learning librarian scraped public posts and created a searchable AI training dataset, without consent of users.