OpenAI’s new image-generation feature for ChatGPT has triggered an explosive reaction, fueling a wave of images that mimic the styles of famous animation studios and cartoonists. As AI-generated visuals resembling real-life individuals surface, concerns over copyright infringement continue to mount. Despite this, OpenAI insists there is no legal issue, while Studio Ghibli has stayed silent and has received no official response.
Similar art styles are possible, but existing artists are blocked
Recently, platforms like X (formerly Twitter) and Instagram have seen a surge of images reimagining current events and historical scenes in a “Ghibli style.” Examples include depictions of President Donald Trump moments after an attack during a 2024 campaign event, along with various altered internet memes. These images are generated using ChatGPT’s image creation feature and are not produced by Studio Ghibli.
Users are posting personal and family photos transformed into Ghibli style. Even OpenAI’s CEO, Sam Altman, changed his profile picture to a Ghibli-style version. ChatGPT can generate a transformed image within a minute when instructed to create a Ghibli-style image, showcasing its user-friendly interface. Numerous images imitating the styles of popular manga, including One Piece, Dragon Ball, and Doraemon, are also circulating widely.
OpenAI explained that when a prompt includes a specific artist’s name, the AI can generate images that resemble that artist’s aesthetic. However, the system is designed to reject prompts referencing contemporary artists to address copyright concerns. OpenAI maintains that Ghibli-style image generation is allowed because it draws on the studio’s overall style, not the individual works of director Hayao Miyazaki.
Studio Ghibli has declined to comment and has shown no indication of pursuing legal action.
Style imitation is not subject to copyright… Is the user legally responsible?
Japanese attorney Satoru Ikemura, an AI and copyright law expert, stated that illustration learning and style imitation are not protected by copyright and are challenging to regulate. Even if an image resembles a specific character, the key issue is how the user uses it, which makes it difficult to hold OpenAI legally responsible.
However, if an AI frequently generates similar images due to repeated learning from a specific character, its operators could bear some responsibility.
Ikemura emphasized the need for technical safeguards that prompt AI to refuse or issue warnings when users input well-known proper nouns. He added that companies should consider adding such features as part of their social responsibility to prevent the inappropriate use of generated images.
