Atlantic investigation reveals millions of songs used for AI music training (www.engadget.com)

An investigation by The Atlantic has unveiled that millions of copyrighted songs, including hits from artists like Taylor Swift and Bad Bunny, have been used to train AI music models. The publication has released four searchable databases totaling over 21 million tracks, providing a comprehensive view into the extent of music exploitation in AI training. This information is particularly significant for the AI/ML community as it highlights the ongoing legal battles surrounding copyright infringement in generative AI, with companies like Suno and Udio defending their practices by invoking fair use. The findings underscore a crucial tension between artistic rights and technological advancement, as many artists and rights holders are increasingly concerned about how their work is being utilized. With existing legal precedents not favoring copyright claims in AI contexts, the extensive databases may empower industry stakeholders to pursue litigation more effectively. As the push for regulation intensifies, especially amid rising incidents of AI-generated music mimicking original content, understanding the scope of used songs is vital for shaping future policies and protections within the music industry.