The Content Overseas Distribution Association (CODA) has countered the latest ambitions of OpenAI by demanding the company halt the use of Japanese-owned content in its new video-generation app Sora, claiming the tool may be infringing copyright by training on anime, games, and other well-known works. CODA says it “has confirmed that a large portion of Sora 2’s output closely resembles Japanese content or images” and argues that the act of training on those works without explicit permission could amount to copyright infringement under Japanese law.
The dispute comes at a fraught time for OpenAI: just as the company has publicly stated its aim to evolve Sora into an “autonomous AI researcher” within a few years, an ambitious goal now shadowed by legal push-back.
OpenAI launched Sora 2 on September 30, 2025, claiming it can handle realistic video generation, sound, and novel scenes with high fidelity. But CODA objects to the use of an “opt-out” system for rights-holders in which content is used unless the owner requests removal, a system CODA says conflicts with Japanese law, which generally requires prior permission rather than retroactive exclusion.
CODA’s letter to OpenAI lists two core demands: that Sora stop using member companies’ works for training without their permission, and that OpenAI respond “sincerely” to copyright-infringement claims from those companies. The Japanese government has also weighed in, warning that Japan’s “irreplaceable treasures”,its manga, anime, and games, must not be undermined by AI models.
For OpenAI, this signals a serious regulatory and reputational risk as it navigates the global AI-training-data landscape. The company may now need to rethink how Sora sources its training data, how it handles generated content, and how it collaborates with rights-owners. Meanwhile, creators and entertainment companies in Japan are watching closely to see if this becomes a precedent for AI’s intersection with intellectual property.
