AI music generator Suno scraped “millions of songs” from YouTube and Deezer

AI music generator Suno scraped “millions of songs” from YouTube, Deezer and Genius to train its AI models, according to hacked data obtained by 404 Media.

A hacker, who goes by the pseudonym Ellie.191, reportedly uncovered data by accessing information from “hundreds of thousands” of Suno users and source code from 2023 and 2024.

According to the data, Suno has used over two million music clips to train its models in creating generative AI music, which it takes from tracks on YouTube Music, podcasts, and lyrics found on sources such as Genius. 

Mixmag clothing capsules: Check out the collection at mixmag.shop 

It also claims that Suno has scraped stock music libraries, including Pond5, Jamendo, Freesound, and the International Music Score Library Project.

Suno has admitted that it trains its models using “all music files of reasonable quality that are accessible on the open internet” in the past, which it says are of “fair use”.

The AI music generator currently faces numerous legal battles, though it recently struck a deal with Warner Music Group (WMG) alongside Udio, another major AI music platform.

Read this next: Artists can now check whether their music has been used to train AI

This week, WMG requested that a US court throw out a case brought against it by a musicians' union over the deal, with claims that it was struck "without compensation or credit" for the musicians that have created the licensed music being used to train it.

Last year, Suno’s CEO, Mikey Shulman, was criticised after claiming that most people "don't enjoy" making music in an interview with 20VC about the future of music and AI.

"I think the majority of people don’t enjoy the majority of time they spend making music," he claimed. "It takes a lot of time, it takes a lot of practice, you have to get really good at an instrument or really good at a piece of production software."

[Via 404 Media]

Gemma Ross is Mixmag’s Associate Digital Editor