A new public search tool from The Atlantic enables musicians and other creators to check whether their work has been included in large-scale datasets used to train artificial intelligence models.
The tool, called AI Watchdog, draws from four datasets that collectively contain millions of songs and are widely shared within the AI development community.
How the Tool Works
One dataset holds 12 million songs, another contains nine million, and two others each have more than 100,000 tracks. Three of the four datasets scraped audio from Spotify and YouTube.
The search function covers all forms of creative work, not just music. Users can look up authors, YouTube creators, screenwriters, and actors.
Researcher Alex Reisner compiled the datasets from research papers and AI data-sharing sites. Companies including Google and Stability have used these datasets.
The tool’s page notes that not every track appearing in a search result has necessarily been used for training. Conversely, a song that does not appear may still be present in a dataset not covered by the tool. Some songs are listed multiple times across the datasets.
Top Artists’ Presence in the Datasets
A search of the tool for the five artists with the highest monthly listeners on Spotify as of June 26, 2026, returned the following numbers of songs:
- Bruno Mars – 133.6 million monthly listeners – 174 songs
- Justin Bieber – 128.4 million monthly listeners – 506 songs
- The Weeknd – 113.9 million monthly listeners – 488 songs
- Rihanna – 110.3 million monthly listeners – 518 songs
- Michael Jackson – 109.6 million monthly listeners – 622 songs
