The OpenAI Studio Ghibli controversy could be a test for art copyright in the AI era

You've probably seen by now that OpenAI's latest AI image generator in GPT-4o can apply a Studio-Ghibli look to anything from common online memes to images from the news. Even brands like McDonald's have jumped on the viral trend of 'Ghiblifying' anything and everything.

People have been using GTP-4o to generate Ghibli-style memes, landscapes, portraits and even events from history and current affairs. While it may seem frivolous, it could be remembered as a turning point for attitudes to AI imagery, and as a key test for copyright.

OpenAI is already facing lawsuits from artists as well as The New York Times, which claims that the company broke the law by training its AI chatbot on copyright material. But the company seems entirely unworried about any potential legal repercussions from the AI Ghibli trend.

Unlike with some previous AI controversies, the Microsoft-backed company is not promising to add new controls to stop its generator from ripping off an artistic style or to stop people from using controversial subjects. If anything, it's actively encouraging the trend, which has renewed public interest in AI image generators.

OpenAI CEO Sam Altman's only concern seems to be the stress it's putting on the company's GPUs as the number of weekly active users rose by 11 per cent. "It's super fun seeing people love images in ChatGPT... But our GPUs are melting," he wrote. He's even switched his profile picture on X to an AI-generated 'Ghiblified' image.

Like some other companies, OpenAI applied restrictions in previous iterations of its AI image generators to prevent imitations of the styles of living artists or the likenesses of real people. but GPT-4o's ability to clearly pastiche the Ghibli style would appear to suggest OpenAI has cast such precautions aside.

"That's just straight up heist of intellectual property – they no longer even bother to pretend its not," one person commented on our initial story.

In a technical paper, OpenAI claims that it's still taking a “conservative approach” to image rights by including a refusal that triggers "when a user attempts to generate an image in the style of a living artist".

But it seems it's now only applying that to individual artists, not studios. So while GPT-4o should, in theory, refuse to generate an image if the prompt asks for the style of Hiyao Mayazaki, it will accept a prompt that asks for the style of Studio Ghibli.

Some users have reported that until the update, the older free version of the AI bot refused to generate images in the style of Studio Ghibli "because it is a copyrighted animation studio, and its artistic style is protected."

The change is curious given that companies can also own copyright, and they're more likely to have the resources to be able to take legal action. It may be that OpenAI believes that it would be more difficult for a studio to prove ownership over a style. Many legal experts argue that in the broadest sense a “style” cannot be copyrighted, so the question will be if the images produced by GPT-4o include specific elements of existing works of art.

So far there has been no official response from Studio Ghibli. OpenAI now has licence arrangements with some companies to allow it to use their content to train its AI models. There has been no statement from either OpenAI or Studio Ghibli about whether this is the case. If there is no agreement, the question that Studio Ghibli will surely want answers to is whether OpenAI trained its model on Miyazaki or Studio Ghibli’s work without permision.

A video has resurfaced in which the great animator and Studio Ghibli co-founded Hayao Miyazaki describes generative AI as an "insult to life itself". Ironically, his work is now prompting the mass mainstream adoption of AI image generators.