Studio Ghibli, Square Enix, Bandai Namco Lead Japanese Crackdown on OpenAI’s Use of Their IP

OpenAI just got a very direct memo from Japan: stop training Sora 2 on our stuff unless we say you can. And yes, that includes some of the biggest names in anime, games, and music.

Who is making the demand

The Content Overseas Distribution Association (CODA) — an industry group that represents 36 major Japanese companies — sent a written request to OpenAI on October 27 and made it public on October 28 (spotted by Automaton). Members span heavy hitters like Studio Ghibli, Square Enix, Aniplex, Toei, and Universal Music. In other words: the people behind a whole lot of the art Sora 2 users would love to generate.

What CODA is actually arguing

CODA says it considers that the act of replication during the machine learning process may constitute copyright infringement.

Beyond the legal phrasing, CODA is pointing to what it sees on the screen: a lot of Sora 2 output looks very close to Japanese content. The group argues that is because Japanese works were used as training data without permission.

The legal wrinkle: Japan is not an opt-out kind of place

CODA’s letter leans on a very specific, very important detail about Japanese copyright law: if you want to use someone’s copyrighted work, you ask first. OpenAI’s current Sora 2 posture — letting companies opt out certain characters after the fact — does not match how Japan handles liability. CODA warns there is no system in Japan that lets you dodge infringement by objecting later.

What OpenAI has said so far

OpenAI CEO Sam Altman tried to get ahead of this in an October 3 blog post. He said Sora 2 will eventually give rightsholders more granular control over character generation, comparable to an opt-in model for likenesses but with extra controls. He also specifically praised Japan’s creative output and said the company wants to share some Sora 2 revenue with rightsholders who actually want their characters generated by users.

CODA’s read: appreciated, but not enough.

What CODA wants from OpenAI now

• Do not use CODA members’ content for machine learning without permission.
• Respond sincerely to member claims and inquiries about copyright issues tied to Sora 2’s outputs, in a way that protects creators while letting AI tech develop responsibly.

That is the whole ballgame: get permission first, and if there is a problem, handle it directly and seriously. Pretty straightforward, even if the tech is not.

One last note from the broader AI mood

In the same general atmosphere of skepticism, Take-Two’s CEO recently said AI is "going to be really, really bad" at making video games and probably could not even come up with the GTA 6 marketing plan. Different company, different corner of the industry, same overall vibe: the people who own the worlds we love are not exactly rolling out the red carpet for generative AI right now.