The music industry faces a widening rift between record labels eager to capitalize on artificial intelligence and the artists whose creative work underpins these emerging technologies. Over the past year, major record labels have partnered with technology companies to develop AI products, yet a fundamental disconnect persists: the musicians whose voices and compositions fuel these systems have largely declined to participate in the initiative, creating tension over who controls the future of music creation.
Universal Music Group, Sony Music and Warner Music Group collectively control millions of songs and theoretically possess the licensing authority to permit their use in AI training. However, the legal and ethical landscape is more complicated. While these labels may own the recorded versions, they cannot unilaterally deploy an artist's voice and likeness for new applications without the consent of the original performer. This distinction has become the battleground in negotiations between labels and technology firms, with musicians increasingly vocal about their refusal to participate regardless of financial incentives.
Prominence has become a liability in these negotiations. Madonna, one of the world's most recognizable recording artists, has made her position unambiguous through her management. According to Guy Oseary, her longtime manager, speaking on the Tim Ferriss podcast, the legendary performer has explicitly rejected any AI involvement with absolute finality. "Madonna does not want her music trained on," Oseary stated. "Don't care what you want to pay her. She's very clear. I do not want my music to be trained on. I want my music to be its own thing." This stance reflects broader concerns within the artist community about surrendering control over their creative identity to undefined technological applications with uncertain commercial viability and reputational consequences.
SZA, the influential R&B artist, has similarly expressed vehement opposition to the practice. When The Atlantic published a searchable database in June revealing which artists' works were used to train popular AI music models without permission, SZA responded with unambiguous language on Instagram: "There's nothing you could ever say to me to make this okay." Her reaction encapsulates the frustration many musicians feel about being incorporated into AI systems without their explicit agreement or understanding of how their voices would be used, reproduced or modified.
Even artists willing to explore AI partnerships are cautious and moving deliberately rather than rushing into agreements. The industry's collective hesitation stems from fundamental uncertainties about financial arrangements and legal protections in this nascent space. Before committing their work to AI training, musicians and their representatives want assurances about compensation structures, mechanisms to prevent unauthorized uses of their voices and likenesses, and clarity about how derivative works would be licensed and monetized. This deliberate approach reflects legitimate concerns about establishing precedents that might disadvantage artists in the long term.
Record labels and streaming platforms, meanwhile, have prioritized demonstrating viable AI strategies to investors, driven partly by market anxiety. Share prices for Universal, Warner and Spotify Technology have experienced steep declines amid concerns about how artificial intelligence might fundamentally disrupt music economics. This investor pressure has accelerated label dealmaking with AI startups. Warner Music Group and Universal have signed agreements with Udio, a platform enabling users to generate music through text prompts. Warner additionally partnered with Suno Inc, which offers comparable functionality with downloadable output capabilities. Universal and independent label representative Merlin are collaborating with Spotify on an AI remix feature.
Remarkably, these announcements were made without securing prior artist commitments. When pressed by analysts about artist participation, label executives offered vague assurances rather than concrete evidence. Michael Nash, Universal's chief digital officer, told analysts on July 30 that the label had engaged in "conversations with thousands of our artists and their estates" and persuaded many to consent, yet declined to name any participants. Warner Music Group CEO Robert Kyncl similarly acknowledged on August 5 that securing artist permission remained laborious and complex, describing it as something "we all have to go through" without demonstrating that this process had been substantially completed.
The legal pathway for labels to deploy music in AI training remains contested. Technically, labels likely possess authority to license music they own for model training without explicit artist approval, though many remain cautious about this strategy given the technology's sensitivity. Some labels have chosen the more diplomatically cautious route of pursuing artist consent even where not legally mandated. Sony Music has taken a notably more restrained approach than competitors, avoiding broad AI partnerships while maintaining active litigation against both Udio and Suno Inc. Warner and Universal had similarly pursued legal action against these same startups before subsequently announcing partnerships with them, illustrating the strategic calculation underlying these maneuvers.
Beyond training datasets, the conflict intensifies when considering how AI companies intend to deploy artist identities. These platforms envision enabling users to generate music by invoking artist names and voices—essentially instructing algorithms to "write a song about a beach day in Taylor Swift's voice." This capability represents a qualitatively different concern than data training, as it directly mimics and commercializes artists' distinctive vocal qualities and public personas. Musicians have demonstrated particular resistance to this application, recognizing that their voices constitute irreplaceable creative signatures that, once digitized and deployed by others, could be used to endorse products, express political views or generate content that contradicts their values or public image.
The unresolved tension between label ambitions and artist resistance reflects broader questions about artificial intelligence governance in creative industries. Malaysian music professionals and Southeast Asian artists face similar challenges as global counterparts regarding how their work might be incorporated into AI systems without consent or compensation. The outcome of negotiations between major Western labels and technology companies will likely establish precedents that influence how regional governments, platforms and artists throughout Asia approach AI deployment in music and entertainment sectors.
