Studio Ghibli Has Few Legal Moves Against OpenAI ChatGPT Image Feature

With ChatGPT’s new Ghiblifying feature, OpenAI is digging into its aggressive stance on copyright law.

OpenAI’s 4o image generation tool, made available for paying ChatGPT users this week, allows people to transform images into the style of Studio Ghibli, the Japanese studio behind anime movies like “Spirited Away” and “The Boy and the Heron.”

If the anime studio brings a copyright lawsuit to stop Open AI from offering the feature — which has been used to make Studio Ghibli-style images of everything from people’s cats to the 9/11 attack on the Twin Towers — it would have a hard time winning in court, copyright lawyers say.

The rules around using Studio Ghibli’s movies to train OpenAI’s models are not yet legally settled, and copyright laws generally allow artists to mimic a visual style, they say.

The Ghibli-style capability, promoted by OpenAI CEO Sam Altman on X, is a step forward from ChatGPT’s previous image generation features.

Earlier versions of ChatGPT, like the one using the DALL-E 3 engine, which is still available in the free version, make it difficult to create images in the style of living artists.

That safeguard is less about copyright protection than about public optics, Matthew Sag, an Emory University law professor studying copyright law and artificial intelligence, told Business Insider.

“OpenAI took a pretty sensible decision to say, ‘We are going to stop producing images in the style of living people,'” Sag said. “Not because it’s copyright infringement, but because people don’t like that. Individuals get very understandably viscerally upset by that.”

On X, OpenAI CEO Sam Altman reacted to the viral Ghiblification by saying the company “put a lot of thought into the initial examples” of its new technology. He wrote the new image-generation features had become so popular that “our GPUs are melting” and he’d impose rate limits while trying to make it more efficient.

Studio Ghibli could bring two kinds of copyright cases against OpenAI

The copyright disputes surrounding ChatGPT and generative artificial intelligence remain hotly disputed in the courts.

Creators have brought dozens of lawsuits across the country against OpenAI and other generative AI companies. On Wednesday, a federal judge in Manhattan allowed the bulk of one of the highest-profile lawsuits — from The New York Times against OpenAI and Microsoft to move forward. Some companies, including Business Insider parent Axel Springer, have struck deals with generative AI companies to allow them to train on their works.

The legal arguments against generative AI companies can be boiled down into two points.

One is the “input” case — that OpenAI would be violating Studio Ghibli’s rights if it were training its large language models on the production company’s movies and TV shows.

The other is the “output” case — OpenAI is creating works that resemble Studio Ghibli’s copyrighted products.

The actual images circulating on social media demonstrate why copyright laws exist in the first place. Studio Ghibli could argue that the wide proliferation of Ghibli-style images — even if it’s as innocuous as people making their children look like they might stumble through a portal into the Spirit World — damages its brand.

Or the company may not want the internet to be filled with Ghiblified photos of the 9/11 terror attacks, or Jeffrey Epstein hanging out with Donald Trump at Mar-a-Lago, or the JFK assassination — all of which have gone viral on X.

The White House even got in on the action, posting a Ghibli-style image of ICE arresting a sobbing woman who was accused of selling fentanyl while living in the US illegally.

An OpenAI spokesperson previously told Business Insider that its policies allow the generation of images in “broader studio styles” but not “the style of a living artist.” But Kristelia García, an intellectual property law professor at Georgetown University Law, said the distinction between copying the work of Studio Ghibli or cofounder Hayao Miyazaki was irrelevant. The studio, and possibly individual artists working for it, could bring a copyright claim, García said.

“You’re still contravening some people who just work the studio,” García said. “So I don’t know. I find that argument to be a little perplexing.”

Representatives for Studio Ghibli and its American distribution company, Gkids, didn’t immediately respond to requests for comment on whether either would sue OpenAI over the feature.

Representatives for OpenAI didn’t immediately respond to a request for comment for this story.

The “input” case remains unsettled in copyright law. And while Japan, where Studio Ghibli is based, has permissive copyright laws for LLM training, the company could still sue in the United States.

But for such an argument to succeed, Studio Ghibli would need to prove that OpenAI’s models were actually trained on the studio’s works, said García. That would likely require a discovery process in the early stages of a lawsuit.

Studio Ghibli would also have a tougher time prevailing in court if it turned out ChatGPT’s LLM was trained on fanart, which has been popular on the internet for decades, García said.

The anime studio would have an even harder time making an “output” argument in court, Christa Laser, an intellectual property law professor at Cleveland State University, said.

In general, while individual works — like Studio Ghibli’s specific movies or scenes or characters — are protected by copyright, artistic style is not.

“If you just evoke the vibe of somebody else’s creative work, it generally doesn’t violate their copyright,” Laser said.

Sag said that, as a visual style, anime is “a cornerstone of AI image generation.”

“You see the distinctive visual language of anime — clean lines, vivid expressions, stylized forms, big eyes, et cetera all over generative AI images,” he said.