The Atlantic created a searchable database of the music used to train AI

The Atlantic has created a searchable public database documenting millions of music tracks found in datasets used to train AI models. Four datasets containing between 100,000 and 12 million songs each have been discovered, downloaded thousands of times, and confirmed to be used by major AI developers in their research. While these datasets are freely available online, developers typically obtain the audio by downloading from YouTube or Spotify using automated tools that bypass platform terms of service, raising questions about how AI training uses copyrighted material. The Atlantic's AI Watchdog site allows the public to search through the songs and other media being used to train AI systems.