Studio Ghibli-backed group urge OpenAI to halt AI training on their content

Studio Ghibli and major Japanese entertainment companies have asked OpenAI to stop training its AI models on their content without permission. The request came via the Content Overseas Distribution Association (CODA), a trade body whose members include anime, manga, and game publishers such as Studio Ghibli, Bandai Namco, Toei Animation, and Square Enix.

CODA issued a formal letter asking OpenAI to confirm that copyrighted works from Japanese creators will not be used to train its AI models—including the Sora video-generation system and future versions. The group stated that Japanese copyright law generally requires permission before works are used for machine learning and highlighted that OpenAI currently relies on opt-out systems.

The request followed the release of OpenAI’s Sora 2 model, after users online shared AI-generated videos resembling anime aesthetics and character styles commonly associated with Japanese studios. CODA said it wants clarification on whether copyrighted work has been included in training data and asked for a framework requiring prior consent.

A Japanese government minister overseeing AI policy also addressed the issue, describing anime and manga as culturally significant and stating that copyright compliance should be maintained in AI development.

Debate around training data and rights has increased globally. In previous reporting, The Washington Post noted that Sora outputs appear capable of reproducing stylistic signatures and known elements, prompting questions about training sources.

Earlier in 2025, a viral trend featuring AI imagery in the style of Studio Ghibli prompted similar copyright-use discussions.

OpenAI has said broadly that it trains on publicly available and licensed data but does not disclose detailed dataset composition. The company maintains mechanisms for creators to request that content not be used for model training.

CODA has requested a response and confirmation that its members’ works will not be used without prior consent. As of the latest reports, discussions between the parties have not been publicly detailed.