- GPT-4o updated includes deeper training on a “variety of image styles”
- Users have created images replicating the hand-drawn look of Studio Ghibli
- Renders raise legal and ethical questions over the work of living artists
Tl;DR What’s the debate?
Fans are using OpenAI’s latest generative model to create images which imitate the distinctive hand-drawn style of Japanese animation house Studio Ghibli. Creatives have questioned whether this is ethical, as artists such as studio founder Hayao Miyazaki are still alive. Legal experts suggest the style itself isn’t protected, but its usage might be.
If you’re a regular on X or Instagram, chances are your feed’s been flooded this week by images that look like scenes from Spirited Away. But these imitation animations aren’t the work of famed Japanese company Studio Ghibli: they’re from ChatGPT. And that has some copyright experts sounding the alarm.
Just a few days ago, OpenAI released the latest version of GPT-4o, the generative AI model which underpins ChatGPT. That update introduced more accurate text rendering, the ability to understand more complex prompts and – crucially – deeper training on a “variety of image styles”.
Within hours of the update, fans were using ChatGPT to create pictures which emulated the art style of different animations. The most common kind of recreation? Renders that bore an uncanny resemblance to the dreamy, hand-drawn aesthetic of Studio Ghibli’s work.
Users have given the Ghibli treatment to everything from popular memes to political figures. But with ChatGPT now apparently able to replicate such an iconic look with ease, there are serious question marks over the legality – and morality – of its usage. Does the Ghibli craze breach copyright? And is it ethical to imitate a living artist? We’ve asked expert lawyers and broken down all of the issues this short explainer.
The debate: users flood social media with AI-generated Ghibli memes
Fan art is nothing new. For as long as there have been animations, fans have created replica artwork. These community creations are usually tolerated by animation houses such as Studio Ghibli, as long as they aren’t used for commercial purposes. Imitation is the sincerest form of flattery, after all.
OpenAI’s latest GPT-4o update changes that. Now, anyone can generate or transform an image into a Ghibli-style scene. What would previously have taken hours can now be achieved in a matter of seconds. Users can emulate art styles with no appreciation of the effort that it would take to a seasoned animator to create the same scene.
Ghibli style memes taking over the internet 😅 pic.twitter.com/96v8ip8ezXMarch 27, 2025
On one hand, this opens up the realm to users who want to create their own fan art but don’t have the requisite skills. On the other, it arguably reduces the value and originality of the source material. Images circulating on social media faithfully replicate the Ghibli look, without a single human artist being involved.
ChatGPT’s ability to channel the visual language of animation houses such as Studio Ghibli makes it a powerful creative tool for fans. But there’s also a moral dimension: some might see such easy imitation as an affront to the studio’s hand-drawn legacy, especially when applied to memes.
Studio Ghibli founder Hayao Miyazaki is known for his painstaking frame-by-frame approach to animation. The studio famously spent more than a year animating a four-second scene in The Wind Rises. In a 2016 documentary, Miyazaki described AI-generated animation as “an insult to life itself”.
Then there are the legal questions. Many of the visuals seen on social media effectively ape the Ghibli look. Because it’s capable of such faithful mimicry, some have questioned whether they infringe on the studio’s copyright.
Awesome. The Ghibli style is now going to become oversaturated and associated with lazy and boring content – can’t wait for kids to grow up thinking the Ghibli movies are Ai-generated and instead of art that’s crafted by excellent artists https://t.co/YFOUIRLYAmMarch 27, 2025
Copyright generally doesn’t protect a visual style as such. So the dreamy backgrounds, oversized eyes and whimsical creatures which are hallmarks of Studio Ghibli animations probably aren’t covered. Unless ChatGPT were to replicate a specific character or scene, it might be hard to prove infringement.
There’s the matter of training data, too. OpenAI hasn’t disclosed specifically what image assets were used to train the latest version of its GPT-4o model. It only specifies that a “vast variety of image styles” featured in the training data. Without more information, it’s impossible to say whether the model was trained on copyrighted Ghibli frames directly – or whether the AI simply learned to emulate the aesthetic from publicly available derivatives, such as existing fan art.
But here’s the verdict today from expert copyright lawyers on where this could all be headed.
The reality: style isn’t legally protected, but its usage might be
We asked Chris Mammen, IP litigator and San Francisco Office Managing Partner for Womble Bond Dickinson, for his expert insight. “Copyright law in the US does not protect ideas or styles,” Mammen explains. “One artist creating a work ‘in the style of’ some other artist has been happening for centuries.”
According to Mammen, AI tools such as ChatGPT are changing two important factors when it comes to fan art creations: “(a) the scale, volume and speed with which new works can be generated, and (b) the fidelity with which these rapidly-generated new works can mimic the style of the original.”
While the style of Ghibli animations might not be covered by copyright, the studio could have other routes to protecting its intellectual property. Mammen continues: “If someone were to represent that one of these GenAI-created images is, in fact, a product of Studio Ghibli, then it could give rise to issues, such as false designation of origin, under trademark and unfair competition laws.”
What’s clear is that AI tools will continue to give rise to new issues around the rules and grey areas involved with derivative and generative content. “These issues are really testing the boundaries of copyright and other intellectual property laws,” Mammen says.
These observations echo those of Matthew Sag, en Emory University law professor studying copyright law and artificial intelligence, who recently told Business Insider: “OpenAI took a pretty sensible decision to say ‘We are going to stop producing images in the style of living people’. Not because it’s copyright infringement, but because people don’t like that. Individuals get very understandably viscerally upset by that,” he said. Studios, though, are a different matter.
Asked for comment by TechRadar, an OpenAI spokesperson said: “Our goal is to give users as much creative freedom as possible. 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. We’re always learning from real-world use and feedback, and we’ll keep refining our policies as we go.”
With tools such as ChatGPT and OpenAI Sora advancing so rapidly, law- and policymakers are scrambling to keep up with the thorny issues arising around intellectual property as a result. As Mammen concludes, “Advocates on both sides of the issue are arguing for legal reform.”
