{"id":1796587,"date":"2026-06-22T16:32:42","date_gmt":"2026-06-22T16:32:42","guid":{"rendered":"https:\/\/celebrity.land\/pt\/?p=1796587"},"modified":"2026-06-22T16:32:42","modified_gmt":"2026-06-22T16:32:42","slug":"music-in-training-sets-is-the-new-spotify-wrapped-how-genai-uses-your-music","status":"publish","type":"post","link":"https:\/\/celebrity.land\/pt\/music-in-training-sets-is-the-new-spotify-wrapped-how-genai-uses-your-music\/","title":{"rendered":"Music in training sets is the new Spotify Wrapped: how genAI uses your music"},"content":{"rendered":"\n
<\/figure>\n<\/p>\n
\n

This has to be peak 2026 music tech. Across every feed, producers I follow are posting the same revelation: they\u2019re finding their own music in training sets used by AI. Investigations by Alex Reisner<\/a> at The Atlantic<\/em> are making this more visible \u2014 and visceral. And the findings are actually much worse<\/em> than I think artists realize.<\/p>\n

The research, the trend<\/h2>\n

The reason you may be seeing this from artists is that The Atlantic<\/em> has made searching the databases simple \u2014 and the information is not paywalled, either. There\u2019s a search box atop Reisner\u2019s investigative series \u201cAI Watchdog.\u201d Enter your artist name(s), and you\u2019re greeted with a list of songs available in various public datasets used for training. These datasets are widely used as the basis for generative music products.<\/p>\n

AI Watchdog<\/a> \u2013 The Atlantic<\/em> (not to be confused with the nonprofit<\/a> of the same name) <\/p>\n

Reisner details the research process and its significance in the full story:<\/p>\n

The Millions of Songs Mashed Into AI-Generated Music<\/a><\/p>\n

(I\u2019m going to overlook their recent editorial record on foreign policy \u2014 like saying maybe killing children is okay<\/a> \u2014 and say, well, 2026 you have to pick and choose and critically filter all media brands to get the whole story. Great work on this one.)<\/p>\n

Not everyone will find their work in datasets. But for those who do, the scale can be astonishing. Berlin-based artist Hainbach has a whopping 151 songs in one dataset alone<\/em>; see his post on the topic<\/a>. <\/em>Here in supposedly privacy-obsessed Germany \u2014 that\u2019s totally false, but that\u2019s another story \u2014 it seems the situation was even worse. From that dataset:<\/p>\n

\n

\u201cA collection of 12,320,916 music tracks from YouTube, totaling 91 years of music. The dataset was assembled by LAION, a nonprofit based in Germany that builds large datasets and has received funding from Hugging Face and Stability AI\u2019s co-founder and former CEO Emad Mostaque.\u201d<\/p>\n<\/blockquote>\n

As Reisner explains, these datasets most often contain just pointers to the music, not the actual audio, but because they\u2019re publicly available, you can bet that some products are using those links to retrieve sound from other audio and video sites. With limited legal precedents or enforcement of usage policies, there\u2019s not much to stop the scrapers. <\/p>\n

So that\u2019s bad, right? <\/p>\n

Well, actually, it\u2019s much worse<\/em> than that. This represents only the publicly available information, shared by researchers. This is the tip of the iceberg. <\/p>\n

Pillaging and looting<\/h2>\n

If your music is not in the search results, for one, don\u2019t be relieved (or disappointed, or however you want to feel). These were readily-accessible examples in this investigation anyone could find by following research footnotes. We know that many products don\u2019t disclose which training sets they\u2019re using. (You should absolutely check and verify the sources.) And then some that do may reveal decisions that unnerve makers.<\/p>\n

Let\u2019s take Google alone \u2014 just one example we know, apart from the bigger iceberg we don\u2019t.<\/p>\n

When Google released Lyria 3<\/a>, they included a statement about responsibilities. Good so far! That includes several measures that should be welcome:<\/p>\n