Leaked Suno code lists 2,013,545 YouTube Music clips (www.theverge.com)

A recent security breach has revealed that Suno, an AI music generator, scraped over two million YouTube Music clips, alongside content from platforms like Deezer and Genius. This leak not only sheds light on Suno's training datasets, which the company has previously kept confidential, but also highlights legal challenges the company faces. The Recording Industry Association of America (RIAA) has been actively pursuing Suno for alleged copyright infringement, arguing that its practices of using copyrighted materials violate legal protections, a claim Suno has contested under the fair use doctrine. The leaked code indicates that Suno employed advanced scraping techniques, possibly using a third-party service called Bright Data to download vast amounts of audio, including a cappella versions of songs. In total, the data suggests that Suno amassed hundreds of thousands of hours of audio content, raising significant ethical and legal implications regarding the use of copyrighted materials for training AI models. Despite the company’s assurance that the compromised source code was outdated and that sensitive customer information remained secure, the incident has sparked concerns about the transparency and legality of AI training practices in the rapidly evolving tech landscape.