A few days ago, SZA denounced that 238 of her songs had been used to train artificial intelligence models without her consent, criticizing especially the exploitation of Black music. Now the Barcelona-based band Medalla claims to have discovered that 14 tracks from their discography appear in one of those same datasets.
In an Instagram carousel that opens with the claim “They’ve stolen from us,” the group explains that a follower alerted them a few weeks ago and that they verified the information through AI Watchdog, the tool championed by The Atlantic that allows searching for works included in various training datasets.
“14 songs by Medalla have been used to train generative artificial intelligence models without our consent. I repeat: without our consent,” they write. The band adds that they don’t know how their recordings ended up there, but they find it “frustrating and dystopian” to witness the repeated violation of artists’ and composers’ rights.
The band also asserts that this is not an isolated case. In a comment on their post, the group notes that they have also found songs from “a lot of friendly bands,” mentioning Camellos, Biznaga, Cala Vento, Carolina Durante, Angelus Apatrida, and La Trinidad.
Medalla emphasizes that if this has happened to a group that is “100% independent, self-managed” and distributes its music from its own label, “who wouldn’t it happen to?” In their message they also warn that this situation threatens to worsen the precarity of the music sector as “techno-oligarchs rub their hands.”
Beyond the complaint, the group urges their followers not to use AI for artistic projects, to financially support independent artists by buying music on Bandcamp, in physical formats or by attending concerts, and to subscribe to newsletters as a direct line of communication in the face of social networks that they view as increasingly unstable.