A coalition of Japan's most powerful animation studios and video game developers has formally accused OpenAI of mass copyright infringement, setting the stage for a landmark legal battle over the future of AI and intellectual property.
In a move that shakes the foundation of the artificial intelligence industry, The Contents Overseas Distribution Association (CODA), a formidable consortium representing the crown jewels of Japanese entertainment, has issued a formal demand to OpenAI to immediately cease using copyrighted properties to train its revolutionary AI video generator, Sora 2.
The association, which includes legendary names like Studio Ghibli, Square Enix, Bandai Namco, and Kadokawa (parent company of FromSoftware), delivered a scathing letter to OpenAI CEO Sam Altman. The core accusation is stark: Sora 2’s ability to generate video is built upon the unauthorized use of their intellectual property, leading to outputs that are direct replicas or clear derivatives of their protected works.
The Heart of the Controversy: From Learning to Replicating
Sora 2, officially unveiled on September 30th with an invite-only iOS release on October 1st, has been both lauded for its technological prowess and criticized for its ethical ambiguity. CODA’s letter leaves no room for interpretation regarding their stance.
“CODA has confirmed that a large portion of content produced by Sora 2 closely resembles Japanese content or images,” the association stated. “CODA has determined that this is the result of using Japanese content as machine learning data.”
The legal argument pivots on a critical distinction. While AI companies often claim "fair use" for training data, CODA asserts that Sora 2 crosses a line. The letter elaborates: “In cases, as with Sora 2, where specific copyrighted works are reproduced or similarly generated as outputs, CODA considers that the act of replication during the machine learning process may constitute copyright infringement.”
This suggests that the AI isn't just learning general concepts like "anime style," but is instead capable of producing content that is substantially similar to specific, copyrighted characters, worlds, and artistic creations owned by its members.
The Legal Chasm: “Opt-Out” vs. Japanese Copyright Law
A significant point of contention highlighted by CODA is the fundamental incompatibility between OpenAI’s proposed solution and Japanese law. OpenAI, like many AI firms, has implemented an “opt-out” system, where copyright holders must proactively request their content be removed from training datasets.
CODA firmly rejected this framework, stating it holds no water in Japan. “According to media reports, it has been stated that Sora 2 responds through an opt-out system based on requests from copyright holders. However, 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.”
This legal stance creates a massive hurdle for OpenAI, potentially requiring it to obtain explicit, prior consent for every piece of Japanese media used to train Sora 2—a logistical and legal nightmare.
The Spark That Lit the Fuse: A Pokémon Problem
The formal complaint was likely spurred by a growing number of examples circulating online that demonstrated Sora 2's ability to mimic iconic Japanese IP with unsettling accuracy. While many instances were reported, one viral video served as a potent symbol of the issue.
A clip, purportedly generated by Sora 2, showed OpenAI’s own Sam Altman playfully running through a field with a group of recognizable Pokémon characters. The caption, “I hope Nintendo doesn’t sue us,” was a joke that now carries a heavy dose of irony. Notably, Nintendo is not a member of CODA, but the incident perfectly illustrated the broader copyright minefield Sora 2 has entered.
What Japan’ Giants Are Demanding
CODA’s demands to OpenAI are direct and unambiguous. The association is not seeking a public debate but concrete action and accountability.
You can read the full, formal letter from CODA to OpenAI on their official website.
The primary demands are:
- Immediate Cessation: OpenAI must stop using the intellectual property of CODA’s member companies without explicit permission for training Sora 2.
- Sincere Engagement: OpenAI must “responds sincerely to claims and inquiries from CODA member companies regarding copyright infringement related to Sora 2’s outputs.”
Failure to comply could see OpenAI facing a series of lawsuits from some of the most litigious and protective rights holders in the world, potentially leading to injunctions that could cripple Sora 2's functionality in one of its key markets.
A Clash of Cultures in a Booming AI Landscape
This legal confrontation unfolds against a complex backdrop in Japan. A recent survey revealed that over half of Japanese companies in the entertainment sector, including major players like Capcom and Level-5, are actively experimenting with AI for game development and other creative processes.
This indicates that the industry is not anti-AI per se, but is fiercely protective of the unique creative IP that forms the backbone of its global success. The message from CODA is clear: technological innovation cannot be built on the uncompensated and unauthorized work of human artists and creators. The outcome of this dispute will undoubtedly set a critical precedent, shaping how AI developers and content creators worldwide navigate the uncharted legal waters of generative AI.