Studio Ghibli and Other Japanese Publishers Demand OpenAI Halt Use of Their Works for Training

A Japanese trade group representing publishers including Studio Ghibli sent a letter to OpenAI last week, urging the AI giant to cease using their copyrighted content to train its AI models without prior authorization.

Studio Ghibli, the acclaimed animation house behind films like "Spirited Away" and "My Neighbor Totoro," has been particularly affected by OpenAI’s generative AI tools. When ChatGPT’s native image generator launched in March, users quickly embraced the trend of transforming selfies or pet photos into visuals mimicking Ghibli’s distinctive aesthetic. Even OpenAI CEO Sam Altman updated his profile picture on X to a Ghibli-style illustration.

Now, with broader access to OpenAI’s Sora video generator, the Content Overseas Distribution Association (CODA) of Japan has formally requested that OpenAI refrain from using its members’ copyrighted works in machine learning without explicit permission.

This appeal is not without merit. OpenAI’s approach—using copyrighted material without prior consent and seeking forgiveness afterward—has made it easy for users to generate deepfakes of copyrighted characters and deceased celebrities. This practice has drawn complaints from entities like Nintendo and the estate of Dr. Martin Luther King Jr., whose likenesses can be readily replicated on Sora.

Whether OpenAI complies remains its choice; noncompliance could lead to lawsuits, though U.S. law currently lacks clear guidelines on the use of copyrighted material for AI training.

Legal precedent is scarce, as U.S. copyright law hasn’t been substantially updated since 1976. However, U.S. District Judge William Alsup recently ruled that Anthropic did not violate the law, despite the company being fined for using pirated books in its training data.

CODA argues that such practices could constitute copyright infringement under Japanese law.

“In cases like Sora 2, where specific copyrighted works are either directly copied or closely replicated in the output, CODA believes the act of copying during machine learning may infringe copyright,” the association stated. “Under Japan’s copyright framework, prior permission is generally required to use copyrighted works, and there is no legal mechanism allowing infringement to be avoided simply by objecting after the fact.”

Hayao Miyazaki, one of Studio Ghibli’s founding creative forces, has not directly commented on the recent surge of AI-generated interpretations of his work. However, back in 2016, upon seeing AI-generated 3D animation, he expressed profound distaste.

“I can’t look at these things and find them amusing,” he said at the time. “I feel strongly that this is an insult to life itself.”