Does OpenAI use book data, such as ebooks or scanned PDFs?

Does OpenAI use book data, such as ebooks or scanned PDFs, to train its artificial intelligence systems?

If so, what sources does it use for this data?
Is it legal for companies to use book data for free to create a business?

Hi,
Check this article to get idea where did training data come from.

This table (blow) from the reference above summaries nicely for those who will not have time to read the blog post: