Over 21 Million Copyrighted Songs Are Circulating Among AI Developers: Report

Four datasets containing more than 21 million copyrighted music recordings are being shared among artificial intelligence developers, according to an investigation published by the The Atlantic.Ad 0:00 Click for sound 0:00 / 0:00 The collections include music from Taylor Swift, Bad Bunny, Billie Eilish, Nirvana, Pearl Jam and the Beatles, alongside legions of independent music producers.

Two of the datasets hold more than 100,000 recordings each while the remaining two are considerably larger, containing roughly 9 million and 12 million tracks, respectively.The Atlantic reported that Google and Stability AI have used tracks from the Free Music Archive, one of the smaller collections.

The largest of the four is LAION-DISCO-12M, a collection of more than 12 million tracks released in November 2024 by LAION, a German nonprofit that assembles open datasets for AI research.The organization is also behind the dataset used to train Stability AI’s Stable Diffusion image generator.

LAION described the music collection as intended for academic use and explicitly warns against deploying it commercially or using it in its original form to create finished products.The dataset contains links to publicly available YouTube tracks and their associated metadata rather than the audio files themselves, and the organization says it does not distribute the music directly.

Each dataset has reportedly been downloaded several thousand times, though the AI industry’s practice of keeping training data confidential means it is largely unknown which companies have relied on which collections.Korean-American producer Kato On The Track, who has crafted hits for Tyga, Snoop Dogg and members of Wu-Tang Clan among other major artists, said 54 of his songs were swept up without consent or compensation.

“Tech companies are using 54 OF MY SONGS to train and sell their generative AI models without compensation or permis...