A Japanese association that reps publishers like Square Enix has called on OpenAI to stop training Sora on Japanese media

I’m sure I don’t need to tell you that a majority of the data that various AI models train themselves on do so without getting any sort of permission to do so. It’s easier to ask for forgiveness than permission, as they say, though in the case of ChatGPT developer OpenAI, they probably shouldn’t have taken that lesson to heart. That’s because Japan’s Content Overseas Distribution Association (CODA) has requested that the AI developer stops using Japanese media to train its text-to-video model Sora.

A touch of context: CODA is a Japanese trade association that represents some of the biggest media businesses in the country. It has a big focus on anti-piracy measures, particularly to do with anime and manga, but in general is about protecting the copyright of Japanese media. You’ll be no stranger to some of the names on the list, as it includes the likes of Bandai Namco and Square Enix, as well as for anime companies like Toei Animation and Studio Ghibli.

Anyway, that request! In a post shared last week (thanks for the spot, Eurogamer), CODA wrote that they have “confirmed that a large portion of content produced by Sora 2 closely resembles Japanese content or images. CODA has determined that this is the result of using Japanese content as machine learning data.”

It goes on to note that Sora 2 operates on an opt-out system, meaning a company has to put in a request if it wants its intellectual property removed from the dataset. The issue is that Japan’s copyright system doesn’t work like that, as you need to get consent before you use any copyrighted works.

CODA’s requests are twofold. First, “In the operation of Sora 2, CODA requests that its members’ content is not used for machine learning without their permission,” and second, “OpenAI responds sincerely to claims and inquiries from CODA member companies regarding copyright infringement related to Sora 2’s outputs.”

I feel it’s important to note that this isn’t saying CODA or any of the companies they represent are anti-AI, and that this is more a concern over the use of their respective bits of copyright. Even before AI, you’ve likely seen fan games and the like receive takedown notices because of their potential to misrepresent a particular work. But something like Sora 2 allows this to be done at scale, so even if you and I know that copyright just exists to protect the interests of big businesses, you can at least see why this is being done.