BTS, BLACKPINK, And More: Thousands Of K-Pop Songs Being Used To Train AI Exposed

As AI continues to expand into more and more aspects of daily life, its various uses have often been debated. Music is no exception, with K-Pop fans calling out many groups for using AI in their releases.

Popular SM Idol Group Fiercely Bashed For Using Too Much AI

This is for various reasons, including environmental impact: data centers require huge amounts of electricity and water, which in turn affect the surrounding areas. How AI learns is also a problem, as seen in a recent development by The Atlantic. 

As part of the media outlet’s recent AI investigation, it found several AI training datasets containing millions of songs.

AI learns by analyzing data, including music, enabling it to create work when prompted. This means that AI, when given the right prompts, could recreate an existing song with slight differences.

As an example, The Atlantic used a pair of figure skaters using an AI song for their performance as an example.

Here is the Czech pair dancing to their AI rip-off of You Get What You Give, replaced for the Olympics by an AI song with “original” lyrics that sounds pretty much the same

— Rodger Sherman (@rodger.bsky.social) 2026-02-10T14:44:15.942Z

In the dataset searching took The Atlantic provided, a search of an artist’s name can show which of the mentioned data sets they are included in. Searching BTS shows that the group’s music, over 200 songs worth in one particular dataset, appears in several of the sets.

Searching other K-Pop groups, especially those with large discographies will show similar results. Of a search of around 11 artists, over 2000 songs were found across the datasets, meaning there are possibly hundreds of thousands of K-Pop songs being used to train AI.

Like the case of the figure skating team, this means that a song created by a K-Pop artist could potentially be plagiarized by AI.

You can read more about a musician’s experience in the video above.