--- title: "A Hack Exposed How AI Music Startup Suno Scraped Millions of Songs" description: "Leaked source code from a hack of Suno, one of the biggest AI music generators, reportedly shows the company pulled millions of songs from YouTube, Deezer, Genius and podcasts to train its models. The revelation lands as Suno fights copyright suits from the major record labels, and after it raised money at a $5.4 billion valuation." category: "Tech" category_url: https://boursel.com/category/tech author: "Rafael Ortiz" published: 2026-07-15T22:25:00.000Z updated: 2026-07-15T22:25:00.000Z canonical: https://boursel.com/article/a-hack-exposed-how-ai-music-startup-suno-scraped-millions-of-songs tags: ["artificial-intelligence", "copyright", "music", "suno"] --- # A Hack Exposed How AI Music Startup Suno Scraped Millions of Songs Leaked source code from a hack of Suno, one of the biggest AI music generators, reportedly shows the company pulled millions of songs from YouTube, Deezer, Genius and podcasts to train its models. The revelation lands as Suno fights copyright suits from the major record labels, and after it raised money at a $5.4 billion valuation. A security breach has pulled back the curtain on one of the most contentious questions in artificial intelligence: what, exactly, went into the model. In the case of Suno, a leading AI music generator, [leaked internal source code reportedly shows the company amassed its training data by scraping millions of songs and hours of audio from major platforms, according to reporting by 404 Media and TechCrunch](https://techcrunch.com/2026/07/15/hack-suggests-ai-music-generator-suno-scraped-youtube-for-training-data/). The details are allegations drawn from stolen code, not admissions by the company, but they arrive at a delicate moment. ## What the leak reportedly shows Suno is a tool that generates original-sounding songs from a text prompt. To build the AI behind it, a model has to be trained on a large body of existing music, and the leaked code reportedly details where Suno's came from. [It points to roughly two million clips, some 113,879 hours, taken from YouTube Music, along with thousands of hours each from the streaming service Deezer and the lyrics site Genius, plus stock-music libraries and about a million hours of audio from around 420,000 podcasts, per 404 Media's account as reported by Variety](https://variety.com/2026/music/news/suno-hack-youtube-music-deezer-genius-data-trained-ai-music-1236811772/). The code reportedly used a commercial scraping service and even included routines to hunt for a cappella versions of tracks. That last detail is legally loaded. Rights holders argue that deliberately working around a platform's anti-copying protections, rather than simply using material that is freely available, strengthens a claim of infringement. ## How it came out, and the fallout The information became public not through disclosure but through a hack. [A breach late last year, using stolen employee credentials, gave an intruder access to Suno's source code, according to 404 Media and TechCrunch](https://www.404media.co/hack-reveals-suno-ai-music-generator-scraped-youtube-deezer-and-genius/). The same breach reportedly exposed customer data, including email addresses, phone numbers and partial payment details. Suno described it as a limited incident that was quickly contained. ## The money and the lawsuits This matters commercially because Suno is not a fringe project. It has raised large sums from investors, [reaching a $5.4 billion valuation, even as it faces copyright lawsuits from the major record companies, as Bloomberg and others have reported](https://techcrunch.com/2026/07/15/hack-suggests-ai-music-generator-suno-scraped-youtube-for-training-data/). Universal Music and Sony Music have been pursuing claims that Suno trained on their recordings without permission; one of the three majors, Warner, has reached a licensing arrangement with the company. Leaked evidence about exactly what was scraped, and how, could shape those cases. ## Why training data is the whole ballgame Underneath the specifics is the central legal and financial question hanging over generative AI. Courts are being asked to decide whether copying copyrighted work to train a model is "fair use," a transformative act that does not require permission, or straightforward infringement. The answer has enormous stakes. If training on protected material without a license is ruled unlawful, AI companies could face large damages and would have to pay to license the data that fuels their products, turning a free input into a recurring cost. If it is deemed fair use, an entire industry gets a powerful shield. A judge weighing that question tends to focus on market harm, whether the AI product undercuts the ability of the original creators to sell or license their work. For AI music specifically, where the output competes directly with human-made songs, that argument is unusually pointed. Suno's leaked code does not settle the law, but it hands the labels something they rarely get: a detailed look inside the machine. Boursel will report how the courts rule rather than predict it. ## Sources - [Hack suggests AI music generator Suno scraped YouTube for training data](https://techcrunch.com/2026/07/15/hack-suggests-ai-music-generator-suno-scraped-youtube-for-training-data/) - [Hack reveals Suno AI music generator scraped YouTube, Deezer, and Genius](https://www.404media.co/hack-reveals-suno-ai-music-generator-scraped-youtube-deezer-and-genius/) - [Suno hack shows how YouTube Music, Deezer and Genius data trained the AI model](https://variety.com/2026/music/news/suno-hack-youtube-music-deezer-genius-data-trained-ai-music-1236811772/)