A viral trend of using generative artificial intelligence to turn photos into illustrations in the style of Japan’s beloved Studio Ghibli has fueled anger from artists and arguments over whether art styles should be protected by copyright.
Starting late in March, X (formerly Twitter) was flooded with images of smiling families on vacation, newlywed couples and other scenes, all in a style closely resembling Ghibli’s My Neighbor Totoro.
This followed OpenAI’s incorporation of a high-quality image generator into ChatGPT, allowing users to produce Ghibli-esque images with just a photo and a simple prompt.
A few days after the update, OpenAI CEO Sam Altman said on X that the service had “added one million users in the last hour.”
Governments jumped on the bandwagon. The White House’s official X account posted a Ghibli-style illustration of a handcuffed and crying Dominican immigrant. The Israel Defense Forces posted images of soldiers carrying guns and piloting a fighter jet, spurring an outcry from fans of Ghibli films accusing the military of using the studio’s style for war propaganda.
Questions have been raised about whether replicating Ghibli’s style so closely constitutes copyright infringement, but there is a broad view that it does not violate current US law.
Those protections are extended to works “fixed in a tangible medium of expression.” For example, specific characters such as Mickey Mouse. But they do not cover ideas or styles. Copying a specific work by a watercolor painter would infringe on their copyright, but creating a new painting in that style would not.
“We continue to prevent generations in the style of individual living artists, but we do permit broader studio styles—which people have used to generate and share some truly delightful and inspired original fan creations,” OpenAI said in a media statement. Studio Ghibli declined to comment to Nikkei on March 27.
This does not mean there is no risk of violating IP rights. Joshua Weigensberg, a partner at law firm Pryor Cashman, has said AI images with features that are “similar or identical to the artist’s works, not only in terms of an abstract artistic ‘style’ but also in terms of particular expressive features, that may also give rise to a legal claim for infringement.”
Trademarks are a potential issue as well, if an AI user or company uses Ghibli-style illustrations in a way that misleads others into thinking that a service is officially tied to the studio.
Creators in the US have decried the Ghibli trend. Digital illustrator Marc Brunet made a post on X criticizing how “Ghibli is making [zilch] out of this while [OpenAI] is raking in the billions after literally stealing their iconic art style.”
Karla Ortiz, an artist and illustrator, has called it a “copyright infringement and a deep violation of artists’ rights and livelihoods.”
The legal landscape for AI is still being established, as are protections for the rights of artists and other creators. Momentum toward barring the use of AI to replicate voices has grown since film star Scarlett Johansson objected last year to OpenAI’s use of a voice “eerily similar” to hers for ChatGPT, and the Ghibli trend may also become a case study for the questions surrounding use of the technology.
“Having it automatically generated, feels much more of a kind of violation of their moral principles and not just of their IP,” said Mark Lemley, a professor at Stanford Law School. “It doesn’t fit within existing copyright law. So courts, maybe, will be motivated to stretch the bounds of existing copyright law to prevent this. Another possibility is that … Congress might change the rules,” he added.
Adobe, a developer of widely used software for digital art, has said it advocates for an “anti-impersonation right” protecting creators against the use of AI to copy their styles. It supported a bill introduced last year by Republican representative Darrell Issa that would ban unauthorized commercial use of digital replicas of individuals, calling it a step toward ensuring responsible use of AI content.
Some believe there should be restrictions placed on the content AI trains on, not just on the AI output itself. That would open OpenAI to a legal quandary if the company had used Ghibli’s IP without permission.
OpenAI has argued that training AI models on publicly available data falls under the fair use exemption of copyright law. The debate on what constitutes fair use in AI applications remains ongoing.
There are roughly 40 US court cases underway revolving around AI training and fair use, according to Lemley. It remains to be seen if the Ghibli-style AI images will continue to take the world by storm.
Japanese studios have “the right to decide whether they want their art copied by generative AI,” Weigensberg said.
This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.