AI-generated music is booming — but at what cost to cultural diversity and creators’ livelihoods? As major deals reshape the industry, experts warn of shrinking revenues, creative homogenisation and unchecked power. “AI can be an opportunity for the whole sector,” they argue, “but only if laws are adapted — and actually enforced.”

At the end of November, Robert Kyncl announced what he called “a landmark pact” — one that, he said, shows artificial intelligence can be “pro-artist” and marks a “victory for the creative community that benefits everyone.” The chief executive of Warner Music Group was referring to a licensing deal the world’s third-largest music company had just signed with AI song generator Suno. Coming just over a week after Warner Music settled a lawsuit and struck a partnership agreement with rival AI music service Udio, the deal remains opaque in its details. According to The Guardian, however, it would allow users “to create AI-generated songs on Suno using the voices, names and likenesses of the Warner acts who choose to opt into the service.” Only last year, both AI song generators had been sued by major music companies, accused of “spitting out millions of AI-generated songs” without artists’ permission. While Universal Music has since reached a settlement, Sony Music is still pursuing legal action against both Suno and Udio.

“These agreements are significant, but also problematic,” warns Giuseppe Mazziotti, coordinator of FAIRMUSE, a European project aimed at promoting fairness in the music industry. “A major concern is whether they can be made without the explicit consent — or even awareness — of the artists involved. If labels license AI companies on the basis of copyright ownership, then the benefits of those licences should also accrue to the creators whose works underpin them.” In Europe, authors’ rights in musical works are represented by collective management organisations. Anyone wishing to use these works must obtain authorisation from the relevant societies, which represent authors and publishers. Yet “when it comes to GenAI services this has not happened”, says Burak Özgen, deputy general manager of GESAC, the umbrella organisation representing authors’ societies across Europe. “On the contrary, our members repeatedly contacted these companies to request cooperation and licensing agreements, and were systematically refused. As a result, apart from some individual deals that major labels had with a few AI company in the US, there are no agreements covering the use of authors’ works for training AI models or operating GenAI services in Europe,” he stresses.


According to a study by the International Confederation of Societies of Authors and Composers (CISAC), the market for AI music solutions is expected to grow by around 60% annually, reaching an estimated value of €16 billion by 2028. Yet a report commissioned by the collective management organisations GEMA and SACEM reveals that 71% of music authors and creators in France and Germany fear that AI-driven music could make it impossible for them to earn a living from their work. “There is indeed a huge gap that needs to be addressed. Value is being created for one side, but that value creation, for now, is coming at the expense of another value that is shrinking,” confirms Özgen, citing further figures from the report: “It can be assumed that by 2028, 27% of music creators’ revenues will be at risk due to unregulated and unlicensed generative AI value chain,” amounting to a cumulative loss that other studies estimate at €22 billion over five years.
“It’sa very risky moment for music creators. If we allow large models to be trained on everyone’s music without proper payment, the entire commercial model of music collapses,” argues Nick Bryan-Kinns, professor of creative computing at the University of the Arts London. “Established musicians who rely on royalties from decades of work would lose that income, and new artists would have no way to get started. If anyone can generate music cheaply with tools like Suno or Udio, why pay emerging musicians?"

Mazziotti shares these concerns, stressing that “FAIRMUSE is grounded in the idea that creators should be able to make a living in digital markets and receive proper recognition for their work, both economically and ethically.” Analysis carried out within the project shows that nearly half of the tracks available on streaming platforms are never listened to. “AI can exacerbate this extreme over-choice by enabling the mass production of tracks whose only function is to dilute the value of professionally created music. The deeper risks lie in broader societal transformations affecting sustainability, cultural diversity and attention,” he emphasises. The preservation of cultural diversity and musical plurality was also among the main concerns of a recently concluded project coordinated by Bryan-Kinns. “The Responsible AI for Music project was about exploring alternative ways of generating music with AI that are less biased towards Western music styles. The problem with current commercial music generation systems is that they are very large AI models, mostly developed in North America, trained on massive datasets dominated by Western classical, rock and pop music. And as AI tools become embedded in music-making workflows, they tend to push creators towards an average, mainstream sound,” he notes.


The risk, he warns, is losing “different rhythms, scales and structures that reflect the richness of our musical heritage, but cannot be generated by these large models.” Yet their unchecked dominance may also threaten the wider economy. “If creators can no longer make a living or lack the resources to continue their careers, the entire cultural and creative industry risks shrinking. Beyond that, culture generates additional value for cities and regions through tourism, transport and consumer services, not to mention the technologies and digital businesses that grow thanks to creative content,” warns Özgen. At the same time, 43% of music authors and creators surveyed by GEMA and SACEM believe that AI “can open up new forms of creativity.” This is precisely what the Responsible AI for Music project demonstrated by testing smaller AI models trained on Indian tabla music, Korean folk traditions, and singing voices absent from mainstream datasets. “Big models like Suno or Udio are extremely easy to use — you just type a prompt and get results. If smaller models are made accessible, they can become part of musicians’ creative toolkits, but the challenge is usability. You need to download them, run them locally, and often have some programming skills. That’s the gap we’re trying to bridge,” says Bryan-Kinns.

Like music authors and creators — more than a third of whom already use AI tools — most music-industry professionals believe the problem is not artificial intelligence itself, but how it is governed. “AI can be a useful tool when applied responsibly and under human supervision,” agrees Mazziotti. “European law is already quite rich in opportunities and rules to counter its harmful effects. An important conclusion of FAIRMUSE is that these laws must actually be put into practice, and that there needs to be greater efficiency and awareness among all the actors involved, starting with EU institutions and both European and national courts.”
Özgen goes further, urging EU lawmakers to introduce a dedicated legal framework for generative AI in the cultural sector, “to fill existing gaps and ensure that all AI services operating in the EU cultural market are subject to EU rules, and creators’ continuous remuneration is guaranteed from the use of AI-generated content in the cultural market, that currently unfairly competes with and substitutes original works, based on which they are developed.” But to truly support livelihoods and cultural diversity, Europe must also “take the power back and invest in European infrastructure for ethical and responsible AI systems,” Bryan-Kinns exhorts. “We can build a flatter, more equitable infrastructure that cherishes cultural heritage and reimagines it for the future. The technology exists. What’s needed now is coordinated strategy and political will,” he concludes.