AI Music Database Exposed
Your favorite artists might have trained AI without knowing it. The Atlantic just launched a tool to search the exact music datasets used by major AI developers. 🔍
What the video says
The Atlantic has launched a searchable database exposing the exact music datasets used to train AI, revealing millions of copyrighted tracks. Reporter Alex Reisner uncovered 4 datasets, including 2 massive collections holding 12 million and 9 million tracks.
While these are freely available online, Reisner outlined how they actually distribute lists of links to YouTube and Spotify. To get the audio, developers use automated tools that bypass logins and ads, violating the platform's terms of service.
Google and Stability are already on record in research papers confirming they used these specific datasets for training. With artists from Lady Gaga to Radiohead logged in the files, anyone can now use The Atlantic's tool to search what has been fed to AI.
