Question 09 · What is AI and how does it work?
Is AI trained with literature?
Short version
Yes. Books are among the most valuable training data, because they contain long, carefully built texts. In some cases, copyrighted works were used without permission, which has led to lawsuits.
Long version
Short internet texts teach a model vocabulary. Books teach it how to develop a thought over many pages, build a character, or make an argument. That is why AI companies have used large collections of books: classics in the public domain, purchased and scanned books, but also collections from illegal online libraries.
In the United States, a court ruled in 2025 that training with lawfully acquired books can be permissible, but using pirated copies is not. The company involved then settled with authors for a sum in the billions. In Europe, the legal situation is less settled.
For writers, a new situation is emerging: their works make AI better, and that AI then competes with them. How they are fairly compensated is one of the major open questions.
AI writes well because people wrote well, and that account is still open.
Sources for this question
- Bartz v. Anthropic PBC, No. 3:24-cv-05417 (N.D. Cal. 2025).
More questions on this topic
Your question is missing?
Send it to frag [at] rogerbasler [.] ch. Selected questions will be answered and added here.