BTS, Psy songs among 60,000 allegedly used by Suno to train AI

BTS, Psy and 2NE1 songs are among more than 60,000 tracks Universal and Sony allege Suno used without permission to train its AI music models.

Published
BTS performs during the Japanese leg of its “Arirang” world tour, held in April.

U.S. AI music generation company Suno allegedly used songs by Psy and BTS without permission to train its AI models, according to documents filed in a U.S. copyright infringement lawsuit.

Songs including Psy's “Gangnam Style” (2012) and “Gentleman” (2013), “Bad Decisions” (2022) by BTS, Benny Blanco and Snoop Dogg and “Waste It On Me” (2018) by Steve Aoki featuring BTS were part of a list of 60,202 songs that Universal Music Group's UMG Recordings and Sony Music Entertainment allege were used in violation of their copyrights. The list also includes “Gettin' Dumb” (2013) by will.i.am, featuring girl group 2NE1.

The list comes from documents submitted to the United States District Court for the District of Massachusetts in September by record labels affiliated with the two music groups as part of a copyright infringement lawsuit against Suno.

The legal battle against Suno began in June 2024, when the three major record companies — Universal, Sony and Warner Music Group — filed a copyright infringement lawsuit against the AI company. Warner later reached a settlement with Suno, while Universal and Sony filed an additional lawsuit in September, seeking $150,000 in damages for each infringed song.

The court has yet to rule on the copyright infringement allegations, but the list was not compiled based on mere speculation by Universal and Sony. U.S. civil litigation includes a discovery process that allows parties, with court authorization, to obtain relevant evidence held by the opposing side.

Under this process, audio experts hired by Universal and Sony were granted access to Suno's security facilities last year and examined the AI company's source code and training data. The experts obtained audio fingerprints from files held by Suno and compared them with recordings owned by Universal and Sony. The analysis produced the list of more than 60,000 songs.

Audio fingerprints are unique identifiers based on the acoustic characteristics of recordings. Music identification services such as Shazam use the technology to recognize songs.

Suno's website

The roughly 60,000 songs do not represent all the music Suno used for AI training. Suno's response filed with the court on Oct. 20 states that “Suno admits that audio data was obtained from YouTube for use as training data using YT-DLP.” YT-DLP is an open-source tool that allows users to download videos and audio from YouTube and other websites.

Suno's data obtained by a hacker also suggests it scraped more than 2 million songs from YouTube, according to a report from U.S. technology news outlet 404 Media. The leaked material included source codes that appeared to be from 2023 and 2024, with a file called “youtube_music” showing that Suno had ingested “2,013,545 music clips.”

In an answer filed in the original 2024 lawsuit, Suno also acknowledged that developing its AI tool required “showing the program tens of millions of instances of different kinds of recordings in order to derive statistical insights about them.”

Suno, however, maintained in the Oct. 20 response that its AI service uses a back-end technological process to create an ultimately non-infringing new product, arguing that the process constituted “quintessential fair use.”

The number of K-pop songs used by Suno for AI training could be far greater than the tracks by BTS, Psy and 2NE1 identified in the recent lawsuit.

(의정부=뉴스1) 권현진 기자 = 가수 싸이가 27일 오후 의정부종합운동장에서 열린 '싸이흠뻑쇼 서머스웨그2026'에서 '강남스타일'을 선보이고 있다. 2026.6.27/뉴스1
Singer Psy performs “Gangnam Style” (2012) at the Summer Swag 2026 concert at Uijeongbu Sports Complex in Uijeongbu, Gyeonggi, on June 27.

The list submitted to the court includes only recordings for which Universal and Sony hold copyrights in the United States or have exclusive rights to manage. Psy's “Gangnam Style” was included in the lawsuit because YG Entertainment previously signed a distribution agreement with Universal covering markets outside Korea and Japan. It remains difficult to determine how many Korean songs AI companies have used for training unless Korean record labels pursue legal action themselves.

Despite these findings, Korean record labels have yet to take any action. Record labels and entertainment agencies were also unaware that songs by Korean artists had been included in the list of allegedly infringed recordings.

“We believe there is a possibility that a significant number of [Korean] musical works have been used for AI training,” the Korea Music Copyright Association, which manages copyrights for Korean musical works, said Tuesday. “We are also considering taking legal action, if necessary, to protect the rights of music creators.”

싸이가 13일 서울 상암 월드컵 경기장에서 열린 '해프닝' 콘서트에서 신곡 '젠틀맨'의 안무를 처음 선보이고 있다.
Singer Psy performs “Gentleman” (2013) during his concert at Sangam World Cup Stadium in Mapo District, western Seoul, on April 13, 2013.

Germany's music rights organization GEMA filed a lawsuit against Suno, alleging that the company used German songs for AI training without authorization. GEMA largely prevailed in the first ruling in July, although the decision came from a German court rather than a U.S. court. The German court rejected Suno's argument that AI training constituted “quintessential fair use.”

GEMA also filed a lawsuit against OpenAI, alleging that the company used German song lyrics for AI training without permission. In an initial ruling last November, a German court largely sided with GEMA.

Authentic Brands Group, which manages rights to some of Elvis Presley's best-known songs, and other rights holders filed a copyright infringement lawsuit against AI company Anthropic last Friday. The rights holders allege that Anthropic used Presley's songs for AI training without authorization.

BY YOON SUNG-MIN [[email protected]]

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.