An anti-piracy organization representing major studios like Square Enix and Bandai Namco isn’t thrilled with how OpenAI appears to have trained its video model, Sora 2.
The Content Overseas Distribution Association (CODA)—a Japanese anti-piracy organization that represents numerous media and video game companies, including Square Enix, Cygames, Bandai Namco, and FromSoftware owner Kadokawa Corporation—has submitted a written request to OpenAI to demand the U.S. company stop using its members' content to train models such as Sora 2.
CODA was formed in 2002 at the behest of the Ministry of Economy, Trade and Industry and the Agency for Cultural Affairs to "promote the overseas development of Japanese content and take anti-piracy measures."
"Japan’s proud content, including music, movies, anime, broadcast programs, video games, and publishing, plays an important role in enhancing the nation’s international presence and economic growth," reads an explainer on the CODA website. "As digital technology becomes more widespread, it is even more significant to protect Japan's content from increasingly artful copyright infringement and promote the content industry’s development."
CODA claims OpenAI might be committing copyright infringement after it confirmed a large number of outputs produced by video generation model, Sora 2, "closely resembles Japanese content or images."
"CODA has determined that this is the result of using Japanese content as machine learning data. 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," reads a statement (first spotted by Automaton).
"Furthermore, 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."
CODA is now requesting that OpenAI prevent its members' content from being used without permission to facilitate the operation of Sora 2. It has also requested the company "respond sincerely" to claims and inquiries from CODA members relating to copyright infringement.
Japanese companies add to chorus of discontent over OpenAI's usage of copyrighted materials
OpenAI launched Sora 2 on September 30, 2025. Just a few days later, company CEO Sam Altman said he wanted to give rightsholders "more granular control" over the generation of characters.
In a blog post entitled 'Sora Update #1,' he said OpenAI intended to let rightsholders specify how their characters can be used by Sora—"including not at all." He also, however, said OpenAI hoped to make the technology so compelling that many people would be happy to see their characters featured.
Notably, Altman also took the time to acknowledge the "remarkable creative output" of Japan. "We are struck by how deep the connection between users and Japanese content is," he added.
OpenAI has been hit with a number of lawsuits in recent years, with many creators taking issue with the company's 'ask forgiveness, not permission' approach to scraping the internet.
Just last week, a judge allowed a consolidated class-action lawsuit from a number of authors, including Game of Thrones scribe George R. R. Martin, to move forward after they claimed OpenAI and Microsoft had violated their copyrights by using their legally protected works to train large language models without permission. Thanks, Business Insider.
OpenAI was candid about the fact that its ChatGPT conversational model was trained on information "publicly available on the internet" as well as "information that we partner with third parties to access" and "information that our users, human trainers, and researchers provide or generate."
It has been less transparent about how Sora 2 was trained. As reported by The Washington Post in September, the company hasn't specified which videos were used to train Sora—although a project co-lead previously explained OpenAI had combined "publicly available and licensed data."
In an attempt to understand what content OpenAI might have used, The Washington Post used Sora to create hundreds of videos that showed the tool is able to replicate Netflix movies and television shows, video games such as Minecraft and Civilization, and popular Marvel characters such as Iron Man and Spider-Man.
Last month, the U.S. Motion Picture Association (MPA) responded to the launch of Sora 2 by reiterating it is the responsibility of OpenAI—not rightsholders—to prevent infringement on Sora 2.
"OpenAI needs to take immediate and decisive action to address this issue," said MPA chairman and CEO, Charles Rivkin. "Well-established copyright law safeguards the rights of creators and applies here."