Studio Ghibli joins other publishers in urging OpenAI to stop using copyrighted works

Japan’s Content Overseas Distribution Association (CODA), which represents leading publishers and studios including Studio Ghibli, has written to OpenAI demanding that the company stop using copyrighted material to train its artificial intelligence models without permission.

Studio Ghibli, the animation studio behind Spirited Away and My Neighbor Totoro, has been particularly affected by AI image generation tools that mimic its distinctive hand-drawn style. Since the release of OpenAI’s image generator in March, users have been prompting the software to recreate selfies and pet photos in a “Ghibli” aesthetic. Even OpenAI CEO Sam Altman briefly changed his profile photo on X to a “Ghiblified” version.

Hayao Miyazaki, one of the central creative forces at Studio Ghibli, has not commented directly on the recent wave of AI-generated imitations of his work. However, when shown AI-generated 3D animation in 2016, he said he was “utterly disgusted.”

“I can’t watch this stuff and find it interesting,” Miyazaki said at the time. “I feel strongly that this is an insult to life itself.”

The letter from CODA follows growing frustration among Japanese media creators about how AI models are being trained. As OpenAI prepares to expand access to its Sora video generator, CODA has asked the company to refrain from using its members’ creative output for machine learning unless explicit consent is granted.

OpenAI has previously faced criticism for adopting what many describe as a “use first, apologise later” approach to copyrighted content. The company’s tools have made it easy for users to generate AI versions of famous characters and deceased public figures, drawing complaints from groups such as Nintendo and the estate of Dr Martin Luther King Jr.

Legal experts say OpenAI could choose to comply voluntarily, but if it refuses, affected rights holders could pursue litigation. Worldwide, copyright law has yet to catch up with AI technology.

CODA, however, argues that such practices could constitute a clear violation of Japan’s copyright laws, potentially setting a legal precedent for AI data training and creative ownership.