The actual content is the clean stuff. If you disagree then you accept OpenAI could just create all the content themselves instead of scraping, which is comparatively trivial.