The Content Overseas Distribution Association (CODA), on behalf of a strong coalition of 36 Japanese entertainment companies, has formally demanded that OpenAI cease using their copyrighted works as training data for its video AI model Sora 2. Well-known brands like Studio Ghibli, Bandai Namco, Square Enix, Aniplex, and Kadokawa are among the member companies.
According to CODA’s letter dated October 28, 2025, a significant portion of the content produced by Sora 2 closely resembles works that belong to its member firms, indicating that the platform may have used their materials for training. CODA contends that OpenAI’s opt-out mechanism is inadequate and unreliable for avoiding liability because Japan’s copyright system requires prior authorization for the use of creative works.
What CODA Is Demanding
The coalition has made two major demands:
- Cease using their content for machine-learning training unless given explicit permission.
- Respond sincerely to claims and inquiries about copyright issues stemming from Sora 2’s outputs.
Japan’s government also weighed in, with ministers affirming manga, anime and game-related IPs are “irreplaceable treasures” and signaling they may tighten regulation of generative AI
This action demonstrates the growing conflict between the rights of creators and the quick development of AI tools in Japan, a nation that exports some of the most influential anime, manga, and video games in the world. This is more than just a reputational risk for OpenAI; it could also result in legal exposure, licensing requirements, or policy changes that have an impact on the development and training of AI models around the world.
Large-scale generative AI models might no longer be allowed to freely draw from copyrighted visual, narrative, and character-driven works—at least not without negotiation or licensing, setting a precedent for publishers and creators.
