Commercial Exploitation of Others’ Works Through Generative AI Alteration
Copyright Dispute Mediations Up Nearly 70% in Three Years; Detection of Infringement Becoming Difficult
"Opt-Out Rights Needed to Prevent AI Usage of Creative Works"

"It was truly chilling to realize that someone had swapped only the faces of the bride and groom just to avoid getting caught. I was shocked at how someone could simply steal another person’s hard work."


Mina Lee, CEO of the studio "Newilmui," recently spotted a strangely familiar wedding photo on a social networking service (SNS). It was a customer photo created by Newilmui, which had handled everything from dress production to hair, makeup, and photography. However, Lee did not initially recognize it as her own work. The first to notice something was off was her manager, who had been working alongside her. After examining the photo closely, the manager remarked, “Even the dress is exactly the same,” tilting her head in confusion. The dress had been meticulously handcrafted by Newilmui. Once they laid the original and the suspicious photo side by side, one similarity after another began to stand out. The pose, the dress, even the small bracelet on the bride’s wrist were identical. The only difference was the faces of the bride and groom.


Another self-studio had taken Newilmui’s client photo and replaced only the faces using generative artificial intelligence (AI), then promoted it as if the studio had taken the picture themselves. Lee explained, “At first, they denied it and claimed it was their own work, but after I presented the evidence, they apologized.” She pointed out, “This is a precious result created together with clients who paid for it, so taking someone else’s work and passing it off as your own should, of course, be punished.”


"My Photo?" AI Swiftly Swaps Faces in Wedding Photos... Copyright Theft Grows Craftier View original image

A new form of copyright infringement is emerging, in which people use generative AI to alter elements of someone else’s photograph or work before passing it off as their own. Unlike previous methods of outright copying, the current approach involves changing only parts such as faces or backgrounds using generative AI to conceal the traces of theft. As it has become more difficult for original creators to recognize that their work has been stolen, there are growing calls for a system to address copyright infringement via generative AI.


According to the Korea Copyright Commission on the 17th, the number of copyright dispute mediation cases increased by 69.9% over the past three years, from 103 cases in 2022, to 110 in 2023, 120 in 2024, and 175 in the previous year. Requests for copyright counseling also rose from 176,171 in 2022, to 142,355 in 2023, 145,997 in 2024, and 175,065 in the previous year.

"My Photo?" AI Swiftly Swaps Faces in Wedding Photos... Copyright Theft Grows Craftier View original image

The steady rise in copyright dispute mediation requests coincides with the growing popularity of generative AI. After ChatGPT was first released in November 2022, Google launched its generative AI service "Bard" in March 2023, followed by Gemini in December of the same year. Throughout this period, dispute mediation requests continued to increase. However, since AI-related disputes are not separately recorded, it is difficult to confirm the exact scale of disputes involving generative AI with current statistics. A representative at the Korea Copyright Commission explained, “We do not categorize AI-related cases separately; instead, depending on the type of copyrighted material, they are counted under categories such as photos or videos.”


The criteria for determining copyright infringement do not change simply because generative AI was used to alter the original. The main questions remain whether the work was based on the original copyrighted material and whether the creative expression of the original is substantially and recognizably present in the result. Therefore, if someone enters a photo into a generative AI tool and swaps only the faces, it can be considered use of the original, and if elements such as composition or subject placement remain intact, substantial similarity is another point for evaluation.


Kim Kiyoon, lead attorney at Kim Kiyoon Law Firm, explained, “Photography is recognized as a copyrighted work since it reflects the photographer’s creative decisions regarding composition, lighting, background, props, and direction.” He continued, “If an individual intentionally alters part of a photo using generative AI and reuses it without the copyright holder’s permission, this can result in both civil and criminal liability for copyright infringement.”


The greater issue is that an environment is developing where, even before legal responsibility is considered, original creators can barely discover that their work has been stolen. In fact, there are increasing cases online where copyright texts or watermarks are removed from photos using generative AI, or only part of the original is altered for commercial use. Kim, who runs a Naver blog, said, “Someone cleverly erased the watermark from my photo, so at first, I didn’t even realize it had been stolen. It was only after other blog users informed me that I was able to report it.”



For this reason, experts argue that, beyond after-the-fact remedies, mechanisms must be put in place at the stage when copyrighted works are used for generative AI, enabling rights holders to retain control. Jeong Sangjo, a professor at Seoul National University School of Law, stated, “It would be a reasonable solution to grant copyright owners the right to opt out of having their works used for generative AI training, while at the same time establishing a text and data mining (TDM) exemption clause that allows certain uses for analysis and training under limited conditions. In cases where opted-out copyrighted works are essential, licensing agreements could be arranged as a solution.”


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing