A data hack reveals Suno emptied the archives of major streaming platforms and lyric sites to train its generative models without permission.
A recent data breach has confirmed that AI music platform Suno built its models by scraping millions of songs and podcasts from YouTube Music, Deezer, and Genius. The leak, reported via 404 Media, exposes source code revealing that over two million clips were ingested into Suno’s training sets between 2023 and 2024.
Beyond major streamers, the platform’s 'fair use' net caught stock libraries like Pond5 and Freesound, as well as the International Music Score Library Project. While Suno claims the open internet is its playground, the industry is pushing back. The company is already entangled in legal disputes over artist rights and copyright infringement.
Despite the friction, major players are hedging their bets. Warner Music Group recently signed a deal with Suno and Udio, prompting a lawsuit from a musicians' union alleging these agreements bypass creator compensation. It’s a cynical move that aligns with CEO Mikey Shulman’s previous claims that most people don't actually enjoy the process of making music.
As the legal battle builds, the breach provides a rare, transparent glimpse into how these models are fed. For electronic music creators, it's a stark reminder that if your audio is public, it's currently being treated as free training data by generative startups.