The Atlantic has turned a quiet corner of the AI industry into something anyone can browse. According to The Verge, Atlantic reporter Alex Reisner recently uncovered four datasets of music being used to train AI models and made them fully searchable for the public.
The scale is striking. Two of the datasets are, in The Verge's words, absolutely enormous — one containing 12 million tracks and another 9 million. The remaining two are described as much smaller, but The Verge notes they still represent a significant amount of music.
Why build a search tool around it? Datasets like these are the raw material that AI music generators learn from, yet they typically sit out of public view. By making the four collections searchable, Reisner gives listeners, and presumably the musicians whose work may be inside, a way to check what has been swept into AI training pipelines.
That visibility lands in the middle of an unresolved fight over how AI companies source the creative material their models depend on, and whether the people who made that material were ever asked.
It matters because, for the first time, the music feeding these systems is something the public can look up by name rather than take on faith.