Some of San Antonio’s most prominent musicians are on high alert after discovering their work was included in a massive dataset used to train generative AI models developed by some of the world’s biggest tech companies.
The datasets were compiled and published in a searchable database by The Atlantic as part of its ongoing “AI Watchdog” series, which examines the books, images, video, audio and other media that tech companies use to train their AI models.
The database consists of four massive datasets reportedly used in AI development. One contains about 12 million tracks, another includes roughly 9 million, while two others each contain more than 100,000 tracks.
The Atlantic cautioned that a work’s inclusion in the database is not definitive proof it was used to train AI models. At the same time, the publication noted that a work’s absence from the database does not mean it wasn’t used for AI training.
Still, the discovery has left some Alamo City artists alarmed, including members of the punk band Girl in a Coma, whose work is easily searchable in The Atlantic’s database.
“Publishing rights are incredibly important, and situations like this highlight why artists need to understand and protect their work,” band members told the Current in a Tuesday report. “We’re not supporters of using AI to create music and believe human creativity should remain at the heart of the art form.”
Both established and rising San Antonio artists have been discovered within the database, including Santiago Jimenez Jr., Temachii, Mypilotis, INOHA, Texas Tornadoes, Flaco Jimenez, Bobby Pulido and Lonesome Rose co-owner and singer-songwriter Garrett T. Capps.
For Capps, the revelation prompted a more tongue-in-cheek response.
“I can’t keep track of all the technological ‘progress’ taking place in these bizarre times. But I will say that I believe in my music, and if robots wanna listen to it and do something with it, then maybe something positive will come from it,” Capps told the Current. “Maybe it will lead to more songs about breakfast tacos? Algorithmic breakfast tacos. Or maybe it will all just lead to the ‘end of the world as we know it.'”
The legality of these tech companies using copyrighted work to train AI models remains an “open question,” The Atlantic reports, although many of them defend their right to do so in court by claiming that training is “fair use,” according to copyright law — even while these models consistently churn out material that’s uncannily similar to work from artists like Michael Jackson, Ed Sheeran and B.B. King.
As of June 2026, musicians and record labels have filed at least a dozen lawsuits against AI companies like Suno, Udio, Google, OpenAI and other vendors for training their models on copyrighted music, The Atlantic reports.