Japanese Publishers Demand OpenAI Stop Training Sora 2 on Copyrighted IP
OpenAI's Sora 2 video generator, launched in October 2025, reportedly generated videos mimicking copyrighted Japanese characters without permission. Japanese content publishers including Bandai Namco and Studio Ghibli demanded that OpenAI stop using their intellectual property to train the model.
What happened
OpenAI launched Sora 2 on October 1, 2025. Users quickly generated videos featuring convincing likenesses of copyrighted Japanese characters including Pokémon, Mario, One Piece, Dragon Ball, and Demon Slayer.
On October 28, 2025, the Content Overseas Distribution Association (CODA), representing Japanese publishers including Bandai Namco, Square Enix, Studio Ghibli, Cygames, Toei Animation, Kadakawa, and Aniplex, submitted a written request to OpenAI. CODA stated it had confirmed that Sora 2's output 'closely resembles Japanese content or images' as a result of Japanese content being used as machine learning data without permission.
Reported impact
CODA stated that Sora 2's replication of specific copyrighted content through machine learning may constitute copyright infringement.
Index analysis
OpenAI CEO Sam Altman acknowledged the popularity of Japanese content-related prompts in his post-launch blog post, stating: 'We are struck by how deep the connection between users and Japanese content is!' On September 29, Reuters reported that OpenAI had contacted studios and talent agencies a week before Sora 2's launch, giving them the option to opt out, though it remains unclear which studios were contacted or whether this included Japanese rightsholders.
Governance controls
CODA made two requests to OpenAI: first, that its members' content no longer be used for Sora 2's machine learning without prior permission; second, that OpenAI respond sincerely to any copyright infringement claims and inquiries from CODA members regarding Sora 2's outputs.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.