Police in Japan have accused a man of unauthorized reproduction of an AI-generated image. This is believed to be the first ever legal case in Japan where an AI-generated image has been treated as a copyrighted work under the country’s Copyright Act.

According to the Yomiuri Shimbun and spotted by Dexerto, the case relates to an AI-generated image created using Stable Diffusion back in 2024 by a man in his 20s from Japan’s Chiba prefecture. This image was then allegedly reused without permission by a 27-year-old man (also from Chiba) for the cover of his commercially-available book.

The original creator of the image told the Yomiuri Shimbun that he had used over 20,000 prompts to generate the final picture. The police allege that the creator had sufficient involvement in the AI image’s creation, and the matter has been referred to the Chiba District Public Prosecutors Office.

Japan’s Copyright Act defines a copyrighted work as a “creatively produced expression of thoughts or sentiments that falls within the literary, academic, artistic, or musical domain.” In regard to whether an AI-generated image can be copyrighted or not, the Agency of Cultural Affairs has stated that an AI image generated with no instructions or very basic instructions from a human is not a “creatively produced expression of thoughts or sentiments” and therefore not considered to meet the requirements to be copyrighted work.

However, if a person has used AI as a tool to creatively express thoughts or feelings, the AI-generated output might be considered a copyrighted work. This is to be decided on a case-by-case basis. The process behind the creation of the specific AI-generated image has to be looked at in order to determine whether it can be considered to be creative enough to be termed a copyrighted work. Key criteria are the amount of detailed prompts, the refining of instructions over repeated generation attempts, and creative selections or changes to outputs.

However, this case is the first of its kind, with no legal precedent. Speaking to Yomiuri Shimbun, a legal expert in AI and copyright from the Fukui Bar Association said that if prompts are detailed and specific enough, then an AI-generated output might be considered a copyrighted work. Basically, the resulting final image is evaluated on how closely it is deemed to match the creator’s original vision, which is determined by how detailed and specific the prompts and their revisions are. “It is vital to determine whether or not the person provided prompts to reach a specific, predicted outcome,” said the attorney.

On the flipside, AI-generated content that closely resembles existing copyrighted works has prompted much furore in Japan lately. This is especially the case with Sora 2 generated video clips that have featured well-known Japanese characters. The Japanese government and an organization representing major creative powerhouses like Bandai Namco, Studio Ghibli and Square Enix have demanded that OpenAI cease unauthorized training on Japanese IPs.

Photographer: Gabby Jones/Bloomberg via Getty Images.

Verity Townsend is a Japan-based freelance writer who previously served as editor, contributor and translator for the game news site Automaton West. She has also written about Japanese culture and movies for various publications.