The Impact of Copyrighted Material on Large Language Models: A Norwegian Perspective
Computation and Language
2025-01-27 v4
Abstract
The use of copyrighted materials in training language models raises critical legal and ethical questions. This paper presents a framework for and the results of empirically assessing the impact of publisher-controlled copyrighted corpora on the performance of generative large language models (LLMs) for Norwegian. When evaluated on a diverse set of tasks, we found that adding both books and newspapers to the data mixture of LLMs tend to improve their performance, while the addition of fiction works seems to be detrimental. Our experiments could inform the creation of a compensation scheme for authors whose works contribute to AI development.
Cite
@article{arxiv.2412.09460,
title = {The Impact of Copyrighted Material on Large Language Models: A Norwegian Perspective},
author = {Javier de la Rosa and Vladislav Mikhailov and Lemei Zhang and Freddy Wetjen and David Samuel and Peng Liu and Rolv-Arild Braaten and Petter Mæhlum and Magnus Breder Birkenes and Andrey Kutuzov and Tita Enstad and Hans Christian Farsethås and Svein Arne Brygfjeld and Jon Atle Gulla and Stephan Oepen and Erik Velldal and Wilfred Østgulen and Liljia Øvrelid and Aslak Sira Myhre},
journal= {arXiv preprint arXiv:2412.09460},
year = {2025}
}
Comments
17 pages, 5 figures, 8 tables. Accepted at NoDaLiDa/Baltic-HLT 2025