Generative AI, Authorship, and the Dialectics of Imitation
The recent case of Chilean producer FlowGPT, who used AI to generate a Bad Bunny–style song, stages an increasingly familiar ethical drama: the clash between creative experimentation and artistic ownership. FlowGPT (Mauricio Bustos) describes his use of generative tools as a way to democratize access to music-making. For Bustos, AI levels the field, as one no longer needs a label, expensive equipment, or even a good singing voice to produce songs that sound professional. It gives the musically imaginative but technically limited creator a chance to enter the arena. This echoes wider discussions of AI music tools, where the promise of democratized creativity sits uneasily alongside unresolved questions of ownership, consent, and artistic control.
At first glance, it appears simple. Using someone else’s likeness — their voice, their sonic identity — without consent seems a straightforward violation. Yet the story unsettles that clarity. Bustos doesn’t hide his method or intent, insisting that his work isn’t an impersonation, but an act of democratization opening the gates of music-making to those without access to studios, voices, or industry networks.
This claim invites a first reversal. What if AI imitation, rather than being theft, could serve as prototype? Like a storyboard in film, it sketches an idea — a sonic demo pitched toward collaboration rather than substitution. Here, AI becomes a way to translate imagination into form. The ethical question then shifts from deception to transparency, from “Is this yours?” to “How are you using it?” At minimum, responsible use hinges on transparency about data and system behaviour, a recurring requirement in ‘human- centred’ AI and ‘responsible’ approaches to generative AI in music.
But even within this generous framing where transparency has morally transformative potential, another tension emerges. The very visibility of such an experiment—its viral life online—depends on the cultural capital of the artist whose likeness it borrows. However transparent the intent, the imitation draws legitimacy from someone else’s earned aura. Here, the line between homage and extraction reappears as AI’s putative openness leans on another’s authenticity. Put simply, is Bustos is feeding on authenticity without replenishing it?
And yet, the audience’s role complicates the accusation because fans like Bustos are not passive consumers. They are also co-authors of the fame they celebrate. Through sharing, remixing, and emotional investment, they help construct the social capital that makes an artist recognizable in the first place. From that angle, FlowGPT’s gesture is as much participatory as it is parasitic, expressing the porous relationship between artist and audience in the age of networked creation.
The dialectic turns again: from ownership to relationality, and from relationality to economy. Every claim about democratization sits inside unequal infrastructures of visibility and reward. AI may widen the circle of creative possibility by lowering barriers to production, but it does not necessarily equalize attention. Digital crowds still gather around the already visible. The technology widens who can create, while filtering who gets heard through the same circuits of popularity, algorithmic amplification, and celebrity capital. This is a familiar pattern: digital tools broaden participation, but attention remains scarce, platform-shaped, and concentrated among a few. Bustos can make music, but it travels because it carries Bad Bunny’s recognizable voice.
What emerges, then, is a dialectical rhythm evident through dynamic tensions between autonomy and participation, originality and dependence, access and appropriation. In that motion of contradiction, AI reveals less about machines and more about us, particularly about how cultural value is produced, circulated, and claimed in an era when voices can be borrowed, remixed, and returned to their makers, sounding both familiar and strange.
