The Japan’s Content Overseas Distribution Association (CODA) has issued a strong warning to OpenAI, demanding that the company immediately stop using copyrighted material from Japanese publishers and studios to train its artificial intelligence models, including its video generator Sora 2.
CODA — an international anti-piracy organization representing entertainment giants such as Studio Ghibli, Bandai Namco, and Square Enix — alleges that OpenAI trained its models on Japanese content without obtaining permission from copyright holders. The group claims that anime, films, and video games from major franchises like Dragon Ball, One Piece, and Final Fantasy were used without authorization.
CODA’s Official Letter to OpenAI
In an official statement, CODA urged OpenAI to immediately cease training its models on copyrighted material belonging to its member companies. The association also asked OpenAI to respond transparently to specific copyright infringement allegations linked to the content produced by Sora 2.
CODA stated that it had confirmed several Sora outputs closely resemble existing Japanese works, suggesting that these may have been included in the training data. The organization also emphasized that, under Japanese law, replicating copyrighted material during the AI training process itself can constitute infringement, even if the content is not publicly distributed.
The “Opt-Out” vs. “Opt-In” Policy Debate
CODA further criticized OpenAI’s earlier opt-out policy, which required copyright holders to actively request that their works not be used for AI training. According to CODA, Japan’s copyright framework operates under an “opt-in” system, meaning creators must grant explicit permission before their works can be used for any purpose — including machine learning.
Although OpenAI has since switched to an opt-in policy, CODA stressed that this does not absolve the company of liability for previous unauthorized use. The association insists that AI developers must obtain prior consent before using protected material for training purposes.
CODA’s Two Key Demands
In its letter, CODA made two formal demands of OpenAI:
- Immediately stop training AI models on copyrighted materials belonging to CODA’s member companies.
- Respond openly and in good faith to any claims and inquiries from Japanese publishers whose works may have been infringed upon by Sora 2.
These demands highlight the growing conflict between AI developers and copyright holders, especially in Japan, where intellectual property laws are among the strictest in the world.
The Japanese Government Steps In
The issue has now attracted attention from Japan’s government. In September, Minoru Kiuchi, Japan’s Minister of State for Intellectual Property and AI Strategy, announced that the government had formally requested OpenAI to avoid any form of copyright infringement. This move signals that the dispute has escalated from an industry concern to a matter of national intellectual property policy.
The confrontation between CODA and OpenAI underscores one of the most pressing challenges in the age of artificial intelligence: how to balance innovation with intellectual property protection. As AI models grow more advanced and capable of generating lifelike creative works, the definition of “fair use” becomes increasingly blurred.
How OpenAI responds to Japan’s publishing industry — and now its government — could set a global precedent for how AI companies handle copyrighted material in their training processes. For now, Japan appears determined to defend its creative industries as technology pushes the boundaries of what machines can learn and recreate.