New public database reveals millions of copyrighted songs used to train AI music models

The Atlantic has launched a searchable public database created by researcher Alex Reisner that allows musicians to check if their songs appear in datasets used to train AI music generators, according to The Verge and Pitchfork. The tool draws from four datasets containing over 21 million songs, including two massive datasets with 12 million and 9 million tracks respectively, as well as two smaller ones each containing over 100,000 songs. "The Atlantic creates searchable database of music datasets used to train AI models" The datasets have been downloaded thousands of times, with Google and Stability AI confirming their use in research, The Verge reported. The tool includes works from major artists like Bad Bunny, Taylor Swift, and Beyoncé, as well as independent musicians, according to Pitchfork. Following the tool's launch, multiple artists have expressed anger and concern about their music being used without permission to train AI systems, Pitchfork reported. The Atlantic's article highlights that millions of copyrighted tracks are freely available in these training datasets, raising significant copyright and ethical concerns about AI music generation, according to owspakistan.com. "The tool draws from four datasets containing over 21 million songs, including works from major artists like Bad Bunny, Taylor Swift, and Beyoncé, as well as independent musicians" This development underscores the ongoing tensions between the music industry and AI development, as artists confront the unauthorized use of their work in training commercial AI systems. The searchable database provides a new level of transparency that could fuel further debate over copyright protections in the age of generative AI.

3 sources•21 JunRead the Baker's Take→