Is OpenAI’s Studio Ghibli-style AI image generator undermining copyright law?

Japanese director Hayao Miyazaki’s anime fantasy films have enthralled generations of children both within Japan and internationally. These movies instilled in countless viewers a lifelong opposition to war, militarisation, and all forms of violence, regardless of their political affiliation. Across Studio Ghibli films like Princess Mononoke, Howl’s Moving Castle, and Castle in the Sky, soldiers are portrayed as misled fighters at best and sexual predators at worst.

This is why there was widespread shock when the official Israeli Defense Forces (IDF) account on X shared images reminiscent of Studio Ghibli’s style.

The Israel Defense Forces used Studio Ghibli’s anti-war art and aesthetics to promote its military | Photo Credit: Israel Defense Forces on X

The IDF’s collage was prompted by OpenAI’s late March update to the GPT4o model. This update enabled users to simply type text prompts into ChatGPT to generate images imitating Studio Ghibli’s iconic animation style.

While most people initially used ChatGPT’s new image update to transform themselves or their family portraits into adorable, Studio Ghibli-inspired images, others ventured further. They created anime stills from photos of the 9/11 terror attacks and the assassination of John F. Kennedy. Users shared Studio Ghibli-style pornography, “Ghibli-fied” stills from Bollywood films, and even a short video clip reimagining Christopher Nolan’s Interstellar as a Studio Ghibli film. What’s more, the irony of Nolan’s Oppenheimer being turned into a Studio Ghibli-style image was lost on far too many.

One AI anime image that gained widespread attention in India depicted the destruction of the historic Babri Masjid in 1992 by Hindutva rioters. This image, rendered in a distorted Studio Ghibli style, featured anime figures waving flags atop the mosque’s dome, cheering in celebration.

Another viral image was shared by the US White House. This picture, in a childish anime style, mocked a foreign citizen accused of drug trafficking who was visibly distraught as she was taken into custody.

Screenshot of the US White House using the Studio Ghibli anime trend to make fun of an arrested person | Photo Credit: The US White House on X

The IDF account on X also joined the trend, sharing anime images of soldiers using communication devices and flying planes.

The art behind the trend

Professor David Leslie, Director of Ethics and Responsible Innovation Research at The Alan Turing Institute, highlighted the ethical concerns surrounding the misappropriation of a man’s life work in “troubling ways.”

“Miyazaki has always been a person who takes a lot of care to present the ideas that are passionate to him in his works, and here you’ve got just an operationalisation of the patterns or the styles of those work[s], now being easily re-appropriated in any context, right? So you’ve got the IDF of Israel producing Miyazaki-style images,” said Leslie.

He described the Japanese artist as “one of the great pacifists of his generation.”

Leslie acknowledged that Studio Ghibli’s media would likely be part of the OpenAI model’s training data set. However, he stressed the need for a more evidence-based approach to explore the extent of its presence. He also highlighted OpenAI’s lack of transparency as a potential complication in this process.

Screenshot of OpenAI CEO Sam Altman reacting to the Indian government’s use of the Studio Ghibli trend | Photo Credit: MyGovIndia on X

Miyazaki’s ardent fans have resurfaced an old video where the Japanese director expressed disgust when introduced to AI-generated motion animation. In the video, Miyazaki referred to AI art as an “insult to life itself.”

Last year, OpenAI championed the fair use defence, which allows copyrighted works to be used in specific circumstances. The company claimed this usage was “fair to creators, necessary for innovators, and crucial for US competitiveness.”

However, Leslie urged the public to consider a broader perspective beyond fair use regulations. He questioned whether OpenAI’s actions were advancing the shareholder value of itself and Microsoft, or if they were aligned with the public interest.

Diving into the legal intricacies, Geoffrey McGovern, Director of Intellectual Property and Senior Political Scientist at the RAND non-profit organisation, clarified that while specific creative products or works enjoy copyright protection, this is not the case for concepts and ideas.

McGovern explained, “For instance, if Georges Braque, the French painter, sought to copyright Still Life with Metronome and the style that became known as Cubism, the painting’s image would theoretically be copyrightable. It’s an image that can be reproduced, and the rights to that reproduction would vest in Braque. However, the ‘Cubist style’ is an idea. US law does not permit the ownership of ideas. It becomes IP-protected only once those ideas are manifested in some form, such as a photograph, poem, recipe, film, composition, and so on.”

McGovern elaborated, “In this manner, the Studio Ghibli style is not copyrightable, and therefore, the ChatGPT-generated image does not constitute a copyright violation.”

Tori Noble, Staff Attorney at the Electronic Frontier Foundation, noted that US courts have yet to determine whether AI models that generate images in an artist’s “style” are considered to be violating copyright.

For instance, in an ongoing case in the US District Court for the Northern District of California, Andersen v. Stability AI Ltd., it was alleged that AI art generators produced images in the style of specific artists. The crux of the matter lies in determining whether the models’ output infringed upon protectable creative expression and whether using works to develop AI image generators constitutes fair use. Noble sheds light on these undecided issues in the case.

Expressing love, sans AI

In our world, there are already works of art inspired by Studio Ghibli that were created without using AI.

One example is The Glassworker, a 2024 fantasy film directed by Usman Riaz. This film was celebrated as Pakistan’s first hand-drawn animated feature film. The character designs clearly demonstrate Riaz’s deep admiration for Studio Ghibli.

Unlike AI-generated images that are hastily assembled by machines in seconds, The Glassworker was meticulously crafted over years. Riaz’s film received expert support from a Studio Ghibli producer and paid homage to Miyazaki’s strong anti-war beliefs. This contrasts sharply with the “Ghibli-fied” images posted by the US White House, the IDF, and individuals who celebrate the destruction of Islamic places of worship in India. This juxtaposition is particularly jarring for those who understand Miyazaki’s anti-war films beyond their superficial appeal as photo filters.

Unlike many civil law countries, the United States does not safeguard the “moral rights” of creators to control how their work is used or adapted, clarified Noble. Consequently, Studio Ghibli films are not entitled to these very limited moral rights in the U.S, she added. Even if applicable, US moral rights would not permit Studio Ghibli to prevent others from imitating their artistic style to create new, transformative works, according to her.

Artists are now contemplating their next steps in response to concerns about their work being illegally stolen or scraped to train AI models.

“This is one of the biggest technology law questions of our age,” said McGovern, “and there is no easy answer that works for all artists.”

He suggested that creators be cautious about posting their work online, or that they explore using paywalls as a safeguard.

Meanwhile, Noble explained that websites take action to limit content scraping through the Robots Exclusion Protocol (REP).

The REP enables websites to include a “robots.txt” file containing rules that restrict data scraping. While these rules are non-binding, OpenAI and other prominent American AI developers currently respect robots.txt restrictions, but not all may, Noble said.

Some artists are utilising anti-AI tools such as Glaze, which are virtually imperceptible to the naked eye but were designed to disrupt or at least mislead AI models attempting to scrape Glaze-protected art for training purposes.

Noble proposed that social and labour policies would be more effective avenues for countries to protect artists and creators from the economic harm caused by AI.

“At least in the U.S., extending copyright protection to uses of works for AI training purposes would not be an effective way to protect artists’ livelihoods, and may actually undermine their interests over time,” she said.

For his part, Leslie cautioned that the viability of living as an artist is not only being challenged by OpenAI but also by other companies that influence how people perceive (or disregard) the work of human artists and art teachers.

“If we kind of widen the lens and really think about the life of artists and the sustainable professional life of artists, the Ghibli moment is really an important but surface-level warning about a deeper transformation that’s happening, where there is, I think, an existential risk to the future [that] visual artists have,” he said.