Large anime, manga and video game companies in Japan have demanded a OpenAI stop using your works without authorization to train your artificial intelligence model Sister 2. The petition was formally presented by the Content Overseas Distribution Association (CODA), an organization that brings together editors such as Aniplex, Bandai Namco, Studio Ghibli, Square Enix, Kadokawa and Shueisha.
CODA announced on October 28 that it sent a written request to OpenAI asking you to stop Sora 2's training with copyrighted material. According to the organization, many AI-generated results “clearly resemble real Japanese works”, which would indicate that the system was trained on copyrighted content, including recognizable characters, visual styles, and scenes.
Possible violation of Japanese laws
In its statement, CODA warned that simply copying works during the machine learning process “could constitute a copyright violation” under Japanese law. Although OpenAI offers an opt-out system for creators, CODA maintains that the country's law requires prior permission before using any copyrighted work. Furthermore, they emphasize that “there is no mechanism that eliminates liability for infringement after a subsequent objection”.
A call for transparency
CODA also asked OpenAI to respond directly to queries from affected companies and clarify whether their model data includes unlicensed Japanese material. For now, OpenAI has not issued an official comment on the claim, while the debate over using copyrighted content to train artificial intelligence continues to grow both in Japan and around the world.
