Home > National > Social Affairs

Ghibli-style AI drawings could be breaking Korean copyright law, legal researchers say

Published: 09 May. 2025, 19:44

LEE SOO-JUNG
[email protected]

Left: Sam Altman, ChatGPT-running OpenAI's chief executive officer. Right: A Ghibli-style image of Sam Altman that was created by generative AI. The image was uploaded as Altman's profile picture on social media platform, X. [YONHAP]

Ghibli-style AI drawings could be in violation of Korean copyright law if the generative AI learned the style with the purpose of producing images, the National Assembly Research Service said earlier this week.

As Ghibli-style or Disney-style images generated by AI become all the rage on users' social media profiles, parliament's research arm chose to provide a study on whether these images would violate the rights of the original artists according to the Korean legal landscape.

A recent Korea Press Foundation poll surveyed 1,000 respondents and found that 59.5 percent said they had used generative AI to transform images into either Ghibli-style drawings or images with specific styles. Of the poll participants in their 60s, 41.4 percent answered that they had used the generative AI's drawing transformation service, showing it has become a fad regardless of age.

Noting that the nondisclosure of data used to train generative AI is causing a legal loophole in judging whether the images violate copyright law, the parliamentary report called for legislation to make AI service providers reveal their data and sources.

In the short term, data used to train AI “should be opened restrictively” at the request of stakeholders, said the research service. To do so, a new provision should be added to the framework act on artificial intelligence.

The Korean parliamentary research agency noted that foreign countries are introducing measures to uncover the data used to train AI.

The European Union has drawn up the Artificial Intelligence Act, which mandates AI service operators to disclose the sources of data used to train AI. The law is set to be effective from August this year, according to the Korean research service.

U.S. lawmakers introduced a bill requiring AI services to submit information about copyright-owned pieces used to train AI.

Yet, the Korean parliamentary research service has taken a cautious approach regarding full disclosure. The parliamentary research service said that full disclosure should be undertaken based on social consensus and development of AI industry. It also noted that copyright law should help “balance the interests” between original artists and AI service operators.

BY LEE SOO-JUNG [[email protected]]