Generative AI, such as ChatGPT, which is rapidly developing and become more familiar, is now facing the question of whether AI's learning of copyrighted work violates the Copyright Act. This is a major issue not only in the U.S., but also in Japan.

The AI Strategy Team, led by the Agency for Cultural Affairs and the Cabinet Office, released a document titled "Relationship between AI and Copyright" on May 30, 2023. The current situation is that the applicable provisions of the Copyright Law distinguish between the AI development/learning stage and the generation/utilization stage. Article 30-4 of the Copyright Law, which was newly revised in 2008, stipulates that, in principle, acts of use that are not intended to enjoy the ideas or emotions expressed in work (i.e., the AI development/learning stage) may be used without the permission of the copyright holders, but the use may be prohibited if it exceeds the "extent deemed necessary" or "would unreasonably harm the interests of the copyright holder, since they are not covered by this provision.

In the generation/utilization stage, the determination of copyright infringement in the case of uploading images, etc. generated work using AI for publication or selling copies is the same as for ordinary copyright infringement. If the generated image is found to be similar to and dependent on an existing image (copyrighted work), the copyright holder may demand damages or an injunction as copyright infringement, and the infringement may also be subject to criminal penalties.

The AI Strategy Team will plan to clarify the interpretation of the Copyright Act (e.g., specific application of "the extent deemed necessary) and promptly disseminate and raise awareness of the current situation through seminars and events. The AI Strategy Team will also study the relationship with future industries such as content industry.

We will follow closely for the latest updates.

Reference Page here (Japanese only): Relationship between AI and Copyright, The Cabinet Office.