On 31 July 2024, the Agency for Cultural Affairs, Government of Japan (the “Agency”) published “Checklist and Guidance related to AI and Copyrights” (the “Checklist”), suggesting some ideas to resolve unsettled issues related to “Do inputs to AI infringe copyrights?” (see our previous blog “Japanese Government Identified Issues Related to AI and Copyrights”) for AI developers as described below:

• When would use of copyrighted works be considered to be done for purposes more than information analysis purposes?

  The following are examples of such cases due to additional purposes to enjoy thoughts or feelings represented on the works:

  • Intentional over-learning.
  • Creation of database containing existing copyrighted works for AI learning purposes to output creative representations of the works through RAG.
  • Additional learning of only a few works of specific creators through LoRA to output creative representations of the works.
• What cases would be considered to “unreasonably harm creators’ interests”?

  Unauthorized use of copyrighted database for AI learning purposes when such database is sold for a fee as an AI learning dataset should be considered so.

The Checklist also shows suggestions for other stakeholders (AI providers and users) to minimize risks related to their own activities on generative AI with use of copyrighted works. The Checklist does not aim to establish any mandatory obligations of, or safe harbor rules for, the stakeholders, but it would be worth following to mitigate risks related to the development and use of AI.

The Checklist further provides guidance to creators of copyrighted works on how to protect and enforce their rights on the works against infringement by those AI developers, providers and/or users. The guidance would be a good source for a creator to get a basic understanding of applicable safeguards and countermeasures to protect its rights on the works.