In April 2023, a track called “Heart on My Sleeve” went quietly viral. The song sounded, unmistakably, like Drake and The Weeknd, their vocal textures, their phrasing, their production aesthetic. But neither artist had recorded a single note of it. It was created by an anonymous producer using AI voice cloning tools, and it racked up millions of streams before being pulled from platforms at the request of Universal Music Group. The music was fake. The legal crisis it triggered was entirely real.
That moment crystallised a question that the music industry, legal scholars, and technologists are now scrambling to answer: when artificial intelligence can clone a voice, imitate a style, and generate a commercially viable song in minutes, who owns it, who profits from it, and who can stop it?
But underneath the legal debate sits a deeper historical reckoning. The unauthorised use of Black artistic voices, their sounds, their styles, their innovations, is not a new problem. What AI has done is automate and industrialise a practice the music industry has been doing to Black artists for over a century.
A History of Taking
Before discussing the technology, it is worth pausing on the history. The story of Black music in America is, in significant part, a story of stolen voice. In the 1950s, rock and roll, a form pioneered by Black artists like Chuck Berry, Little Richard, and Fats Domino, was covered by white artists who outsold the originals and collected the cultural credit. Elvis Presley’s versions of songs like “Hound Dog,” first recorded by Big Mama Thornton, generated fortunes that largely bypassed their Black originators. Radio programmers and record labels systematically promoted white-performed versions of Black music to white audiences.
This was not coincidence. It was industry practice. The term “race records,” used by labels to categorise music by Black artists, reflects the segregated logic of a market built on Black creativity that directed its revenues toward whiteness.
AI voice cloning does not replicate this history accidentally. It inherits it. When the first widely discussed AI voice controversies involved Black artists, Drake, The Weeknd, Kendrick Lamar, or when AI tools are used to generate music “in the style of” Black genres like hip-hop, R&B, and soul, the technology is reaching into the same cultural reservoir it always has. The difference is speed, scale, and legal ambiguity.
How AI Music Generation Actually Works
To understand the legal tangle, it helps to understand the technology. Modern AI music tools, including systems like Suno, Udio, and various voice cloning platforms, are trained on vast datasets of existing music. They learn patterns: what makes a melody feel resolved, what distinguishes a hip-hop beat from a bossa nova rhythm, what phonetic and tonal characteristics define a particular singer’s voice.
Given a text prompt, “an aggressive trap beat with a gravelly Southern rap vocal,” they generate something new that reflects those learned patterns. The output can be remarkable. It can also be legally murky in at least three distinct ways: the training data may have included copyrighted songs without permission; the output may closely resemble a specific artist’s voice or style; and questions about who owns the output remain unresolved in most jurisdictions.
The Training Data Problem
The first battleground is the training data. Most major AI music platforms have been trained on millions of songs, the vast majority of which are protected by copyright. The AI did not listen to those songs for enjoyment, it used them to build a statistical model of what music sounds like.
Music industry lawyers argue this constitutes copying. AI companies argue it constitutes transformative use, analogous to a human musician learning from listening to records. But there is a critical distinction: the human musician who learns from Marvin Gaye’s What’s Going On then creates something genuinely new, shaped by thousands of other influences and personal experience. An AI system produces outputs that are statistically derived from its inputs, a fundamentally different kind of “learning.”
In 2024, Universal Music Group and other major labels filed suit against Suno and Udio, alleging that those systems were trained on copyrighted recordings without licences. The catalogues at the centre of these suits include decades of Black music. If AI companies can train on that music without compensation, they are, in effect, extracting value from Black creativity for the second time, first when labels undercompensated original artists, and again when AI developers used those recordings without payment.
The Voice Cloning Crisis
Separate from training data is the specific question of voice cloning, the ability to generate audio that sounds like a specific, identifiable person. This is where the ethical stakes feel most immediate, and where the resonance with Black history is most acute.
Voice is deeply personal. For Black artists specifically, the distinctiveness of voice has been both a source of power and a target of appropriation. The specific qualities of a voice, its timbre, its inflection, the way it carries cultural memory, are not abstract sonic properties. They are the product of lives, communities, and histories.
When an AI system clones Drake’s voice without consent, it is doing technologically what record labels did contractually for generations: extracting the value of a Black artist’s distinctive identity and redirecting it. The mechanism is different. The logic is the same.
Some jurisdictions offer limited protection through rights of publicity laws, which prevent commercial exploitation of a person’s name or likeness without consent. Tennessee, home of significant music history including a large Black musical heritage, passed specific AI voice protection legislation in 2024. But global coherence is far off, and enforcement is even further.
The Style Appropriation Question
Copyright law has historically not protected musical style, only specific compositions and recordings. This creates a significant gap when AI is involved. An AI system trained on decades of Black music can generate endless tracks “in the style of” hip-hop, gospel, blues, or soul without technically infringing any specific copyright. The style, the cultural property of entire communities, is not protectable.
This mirrors a longstanding grievance in music law. When Led Zeppelin took structural and stylistic elements from Black blues musicians, they often did so legally. When the Beach Boys built on Chuck Berry’s chord progressions, they paid no royalties to Berry’s community. Style has always been a vector for extraction, and AI makes that extraction faster and cheaper.
Some legal scholars and advocates are beginning to argue for a broader concept of cultural rights, protections that go beyond individual copyright to recognise the communal ownership of cultural forms. This is a significant doctrinal shift, and it faces enormous resistance from industries built on the free use of cultural influence. But the AI moment is making the argument more urgent.
What Artists Are Doing
Some Black artists are fighting back. Jay-Z’s company Roc Nation has been active in challenging unauthorised AI uses of his voice. The estates of deceased Black artists, from Tupac Shakur to Aaliyah, have been vigilant about AI-generated material that uses their voice likenesses.
Others are leaning in on their own terms. Timbaland launched an AI music collaboration platform designed to allow artists to maintain control of their sound while experimenting with generative tools. The key distinction is consent and compensation, not the technology itself, but who controls it and who benefits.
The Deeper Question
Behind every specific legal question is a larger one rooted in history: what happens when a technology makes it effortless to replicate a culture it did not create? Black music has always been the engine of American and global popular culture. The question of whether AI will finally force an honest reckoning with that legacy, or simply become the latest instrument of its exploitation, will be answered not just in courtrooms, but in the choices that technologists, executives, and listeners make right now.
The law will catch up, eventually. In the meantime, the music keeps playing, and the rules are being written in the air.
