OpenAI’s Sora 2 is back in controversy as some of Japan’s most powerful entertainment companies are taking a stand against the platform. The group, represented by the Content Overseas Distribution Association (CODA), has formally demanded that OpenAI stop using its copyrighted works to train its AI. CODA’s members include some of Japan’s most recognised names in anime and gaming, Studio Ghibli, Bandai Namco, Square Enix, Toei Animation, Kadokawa, and Aniplex.
These companies argue that a large amount of Sora 2-generated content mirrors the look and feel of iconic Japanese creations, indicating that their works were used in training without consent. Under Japanese copyright law, using copyrighted material requires prior approval, and there’s no legal basis for post-use opt-out systems. Yet, OpenAI’s current policy allows creators to exclude their content only after it’s been used, a setup CODA says violates Japan’s strict IP protections.
Interestingly, the Japanese government has also stepped in, calling many of the affected works “irreplaceable cultural treasures.” Officials have urged OpenAI to immediately stop infringing on Japanese art and cultural content. The pushback follows an online surge of AI-generated videos that resemble popular Japanese franchises, from Pokémon and Mario to One Piece, Dragon Ball, and Demon Slayer.
OpenAI CEO Sam Altman has acknowledged the growing criticism and said the company plans to revise its opt-out system to give rights holders more control, potentially moving toward an opt-in model. However, the issue highlights a deeper global debate over whether AI companies can legally use copyrighted content to train models without explicit permission.
For Japan’s entertainment giants, the message is clear: their art, characters, and stories are not free training material. And for OpenAI, Sora 2’s future in one of the world’s most influential creative markets could now depend on how it responds to these calls for accountability.
