黑客泄露的数据显示，AI 音乐生成公司 Suno 通过从 YouTube Music、Deezer 和 Genius 等在线平台抓取数百万首歌曲和歌词来训练其模型。
这是外界罕见地一窥 Suno 训练数据集的实际来源，此前该公司一直回避披露训练数据内容及获取方式。 该数据泄露正值 Suno 面临多起诉讼之际，其中包括美国唱片业协会（RIAA）提起的、指控其使用受版权保护材料训练 AI 模型的案件。
在一次黑客事件中获得的Suno数据曝光了，AI音乐生成器是通过从包括YouTube Music、Deezer和Genius在内的在线音频平台抓取数百万首歌曲和歌词进行训练的，404 Media 报告：https://www.404media.co/hack-reveals-suno-ai-music-generator-scraped-youtube-deezer-and-genius/。鉴于Suno一直未透露其训练数据集的内容及获取方式，这提供了一个罕见的机会来看Suno实际上从在线平台获取了什么。
黑客（被称为“ellie.191”）向404 Media分享的材料据称证实了这些指控。数据包括2023年和2024年的Suno源代码，以及从YouTube Music、Deezer、Genius、Pond5、Jamendo、Freesound和国际乐谱图书馆计划（IMSLP）抓取音频文件的指南。其他泄露的代码据称显示，Suno曾使用一家名为Bright Data的第三方公司从YouTube抓取音乐，并似乎在该平台上搜索歌曲的无伴奏版本以获取仅有人声的音频。
一个针对YouTube Music的文件显示，在其最后更新时，Suno已处理了2,013,545个YouTube Music剪辑。根据另一个文件，Suno编制的数据集包括数十万小时的YouTube Music，数千小时的Deezer、Genius、IMSLP、Jamendo和Pond5，以及数百小时的Freesound和MuseScore歌词。根据额外的代码，Suno还计划通过名为PodcastIndex的在线工具下载约一百万小时的播客。
一位未透露姓名的Suno发言人在向404 Media发表的声明中表示：“正如我们在公开文件和披露中所述：https://help.suno.com/en/articles/9709569?ref=404media.co，Suno的AI模型是基于公开可用的音乐文件及可在开放互联网第三方网站访问的相关元数据进行训练的。”
黑客还访问了Suno的客户信息，包括电子邮件地址、电话号码和Stripe支付详情。一些接受404 Media联系的客户确认他们确实注册了该服务，并表示Suno从未通知他们发生过安全漏洞。
在向404媒体发表的一份声明中，Suno的一位发言人表示，公司在2025年11月意识到一起安全事件，并且情况很快得到了控制。
Suno data obtained in a hacking incident has exposed that the AI music generator was trained by scraping millions of songs and lyrics from online audio platforms, including YouTube Music, Deezer, and Genius, 404 Media ：https://www.404media.co/hack-reveals-suno-ai-music-generator-scraped-youtube-deezer-and-genius/ reports. Given that Suno has avoided revealing what’s in its training datasets and how they were acquired, this a rare glimpse into what Suno has actually been taking from online platforms.
Materials shared with 404 Media by the hacker, referred to as “ellie.191,” reportedly back up those allegations. The data includes Suno source code from 2023 and 2024, alongside scraping instructions to pull audio files from YouTube Music, Deezer, Genius, Pond5, Jamendo, Freesound, and the International Music Score Library Project (IMSLP). Other leaked code reportedly suggests that Suno used a third-party company called Bright Data to scrape music from YouTube, and seemingly searched for a cappella versions of songs on the platform to source vocal-only audio.
A file for YouTube Music notes that Suno had consumed 2,013,545 YouTube Music clips at the point it was last updated. According to another file, datasets compiled by Suno included hundreds of thousands of hours of YouTube Music, thousands of hours of Deezer, Genius, IMSLP, Jamendo, and Pond5, and hundreds of hours of Freesound and MuseScore lyrics. Suno also sought to download roughly one million hours of podcasts via an online tool called PodcastIndex, according to additional code.
“As we have stated in public filings and disclosures：https://help.suno.com/en/articles/9709569?ref=404media.co, Suno’s AI models have been trained on publicly available music files and related metadata accessible on third-party websites on the open Internet,” an unnamed Suno spokesperson said in a statement to 404 Media .
Suno customer information was also accessed by the hacker, which included email addresses, phone numbers, and Stripe payment details. Some of the customers contacted by 404 Media confirmed that they had signed up for the service, and said that Suno never notified them about a security breach.
In a statement to 404 Media , a Suno spokesperson said the company became aware of a security incident in November 2025, and that the situation was quickly contained.
扫码直达文章详情