October 18, 2023
TOKYO – The Japan Newspaper Publishers and Editors Association called for an early revision of the Copyright Law — which currently allows, in principle, the unauthorized use of copyrighted material for machine learning — at a meeting of a Cultural Affairs Agency body on generative AI and copyright.
The law’s Article 30-4 — newly established in a 2018 revision — allows artificial intelligence systems to machine-learn copyrighted works, such as newspaper articles, without permission from the copyright holders. The law makes an exception for cases that “would unreasonably prejudice the interests of the copyright owner” but provides few specific cases.
At Monday’s hearing, held by a subcommittee of the Culture Council, the newspaper association, also known as Nihon Shinbun Kyokai, presented an example of a generative AI-equipped search engine using news articles — including ones meant to be limited to subscribers — to answer online questions. The association claimed that a large portion of the AI responses to the questions was generated based on newspaper articles learned without permission and that the AI presented a number of expressions similar to the original articles.
Citing the fact that newspaper companies sell collections of article data for information analysis, the association stressed that machine learning of articles “unreasonably prejudices the interests of the copyright owner” as described in the wording of the law. It then stressed the need to revise the law and establish a system that allows copyright holders to bar machine learning or to be asked for permission at the time of learning.
Researchers from a national AI research and development corporation who were at the hearing said that they would conduct research and development that would contribute to the prevention of copyright infringement, based on technical issues and legal discussions.