Japan's CODA urges OpenAI to stop using member content to train models like Sora 2

• CODA says many Sora 2 results closely resemble Japanese works and imagery.
• Member companies include Studio Ghibli, Bandai Namco, Square Enix, and Shueisha.
• CODA argues Japan’s copyright law requires prior permission, not post-use opt-outs.
• The group demands that OpenAI respond to infringement claims from its members.

Japan’s Content Overseas Distribution Association (CODA) has asked OpenAI to stop using its members’ content to train models like Sora 2.

As reported by Automaton, the association alleges OpenAI may be infringing copyrights, saying many Sora 2 outputs closely resemble Japanese content or imagery.

CODA said the request was made at the urging of its member companies, including Aniplex, Bandai Namco, Studio Ghibli, Square Enix, Kadokawa, and Shueisha.

The association claims that many of Sora 2’s outputs resemble Japanese works, suggesting that the model used Japanese content during training. It warns that such copying during machine learning may amount to copyright infringement.

“Under Japan’s copyright system, prior permission is generally required for the use of copyrighted works, and there is no system allowing one to avoid liability for infringement through subsequent objections,” CODA said in a statement. 

Permission disputes

CODA has asked OpenAI to stop using its members’ content without permission for Sora 2 and to respond sincerely to copyright infringement claims and inquiries from its members.

While Sora 2 allows copyright holders to opt out, CODA argued that Japan’s laws require prior permission to use copyrighted works and that post-use objections do not exempt liability for infringement.

Over the past few months, OpenAI has faced multiple lawsuits from creators accusing the company of using their work without consent.