AI Is Learning From Musicians. Who Gets to Profit?

Generative AI is changing how music is made, but the recordings, compositions and performances behind these systems belong to real artists. As AI-generated music enters the industry, musicians are questioning who gave permission, who gets paid and who controls what happens next.

Music has always been shaped by what came before it. Artists borrow ideas, respond to influences and build on the sounds of earlier generations. That exchange is part of what makes music evolve.

Artificial intelligence introduces a very different proposition. Rather than drawing inspiration from individual records or artists, generative AI systems can analyse vast quantities of existing music and use what they have learned to produce new recordings, compositions and vocals.

Much of that source material belongs to musicians who may never have agreed to its use. They may not know whether their recordings have been included in training datasets, have no opportunity to object and receive no payment when the technology is used commercially.

And the implications go beyond how music is created. AI-generated acts are entering the market, competing for listeners and, in some cases, attracting audiences that established musicians have spent years building.

The question is no longer simply whether AI can make music. It is whether the industry is prepared to let technology companies build commercial products from musicians' work without giving those musicians meaningful control over how it is used.

The numbers behind the concern

A report published in January 2026 by the Independent Society of Musicians, the Society of Authors, Equity, the Association of Illustrators and the Association of Photographers provides some of the clearest evidence yet of the impact generative AI is having on creative work in the UK.

Brave New World? Justice for Creators in the Age of GenAI brings together evidence from more than 10,000 creators across music, writing, photography and performance.

Its findings on musicians are particularly concerning.

According to the report, 53% of surveyed musicians said they had already lost work to generative AI, or could not be sure whether they had. Separately, 73% said unregulated generative AI threatened their ability to earn a living.

The report also describes musicians experiencing income cuts of 40–50% as paid briefs dry up. Session work and songwriting opportunities are among the areas affected, with studios reportedly replacing musicians with AI-generated alternatives.

There is another figure that deserves attention. Only 7% of surveyed musicians said they had ever been approached to license their work for AI training. Fewer than one in five of those approached had received payment.

That raises a fundamental question about the commercial use of music. If existing recordings, compositions and performances are helping to develop systems capable of generating new music, why are so few of the musicians behind that source material being approached about licensing it?

The report covers other creative professions too. It found that 32% of surveyed illustrators had lost commissions to AI-generated art, while 58% of professional photographers reported losing assignments to generative AI by February 2025. Among authors, 86% said generative AI had already reduced their earnings.

These findings are based on surveys and reported experiences rather than a definitive count of jobs lost solely because of AI. But they demonstrate that the economic effects are already being reported across creative professions.

For musicians, the concern is not simply whether AI can generate a convincing song. It is what happens when that music enters the market, competes for audiences and potentially takes opportunities away from artists trying to make a living from their work.

When the competition is built from your own work

The experience of Welsh post-hardcore band Holding Absence offers a particularly uncomfortable example.

In October 2025, frontman Lucas Woodland spoke out after an AI-generated project called Bleeding Verse overtook Holding Absence in monthly Spotify listeners after just two months.

At the time, Bleeding Verse had 897,349 monthly listeners, compared with 847,638 for Holding Absence. The project described itself as being inspired by artists including Holding Absence and Dayseeker, using AI-assisted instrumentation and vocals.

Woodland described the experience as shocking, disheartening and insulting. He called it a wake-up call and urged people to oppose AI music, warning that bands like his could stop existing.

The significance of the example goes beyond a comparison of monthly listener figures. Holding Absence had spent years developing its sound, releasing records, touring and building an audience. An AI-generated project openly citing the band as an influence was able to attract a larger monthly listener count in a matter of months.

It raises difficult questions about originality, ownership and the value of creative work. If an AI-generated act can reproduce recognisable characteristics of a band's music and attract listeners, where does that leave the musicians whose work helped establish those characteristics in the first place?

The distinction between influence and imitation matters here. Musicians have always drawn inspiration from other artists. Entire genres have developed through musicians responding to, borrowing from and reinterpreting the work of those who came before them.

Generative AI changes the scale and commercial context. Systems can analyse vast catalogues of existing music and produce new material based on what they have learned. The artists whose work informs those systems may have no meaningful say in how it is used, or whether they receive anything in return.

Monthly listener figures are not a measure of income or proof that one act has directly displaced another. But the example does expose an uncomfortable possibility: musicians could find themselves competing against projects that draw on the sounds and identities of the very artists who spent years creating them.

What are music platforms doing?

Music platforms are beginning to introduce policies around AI-generated content, but their approaches differ considerably.

Bandcamp has taken the clearest position. In January 2026, it announced that music and audio generated wholly or in substantial part by AI would not be permitted on the platform. It also prohibits the use of AI tools to impersonate other artists or imitate their styles.

Its policy is rooted in the idea that music is a human cultural activity and that fans should have confidence in the music they discover on Bandcamp. Users can report suspected AI-generated material, which the platform reserves the right to remove.

Deezer has taken a different approach, developing technology to identify fully AI-generated recordings. It tags detected tracks, removes them from algorithmic recommendations and editorial playlists, and demonetises streams identified as fraudulent.

By January 2026, Deezer reported detecting and tagging more than 13.4 million AI-generated tracks during 2025. It was also receiving around 60,000 AI-generated tracks every day. The company has made its detection technology available to other industry players, arguing that the issue requires a wider response.

Spotify has chosen not to ban AI-generated music. Instead, its focus is on protecting artist identities, tackling spam and giving listeners more information about what they are hearing.

It has strengthened its rules around unauthorised AI voice cloning, introduced AI disclosure credits that allow artists and rights holders to identify where AI has contributed to vocals, lyrics, instrumentation or production, and developed measures to prevent music being fraudulently uploaded to another artist's profile.

In August 2026, Spotify also announced AI Persona badges for profiles representing AI-generated identities. These badges identify artist personas that do not represent real people, and music from those profiles is excluded from editorial and personalised recommendations by default. The badge is about the identity behind the profile, rather than whether an individual recording uses AI.

TIDAL has introduced another approach. It permits AI-generated music, but recordings it identifies as wholly AI-generated are labelled and are not eligible for royalty attribution. It can also block or remove content associated with fraud, impersonation or attempts to interfere with genuine artists and their audiences.

The differences are significant. Bandcamp has drawn a line around what it will host. Deezer is identifying and restricting AI-generated material, particularly where it is associated with fraudulent activity. Spotify is concentrating on disclosure, impersonation and artist identity, while TIDAL has gone further by withholding royalties from recordings it identifies as wholly AI-generated.

These are meaningful steps, but they address what happens once AI-generated music reaches a platform. They do not necessarily address how the systems producing that music were trained in the first place.

Labelling an AI-generated track does not resolve whether the music used to train the system was licensed. Removing fraudulent streams does not compensate the musicians whose recordings may have been used. Preventing an AI-generated act from impersonating a singer does not necessarily protect that artist's wider creative identity.

There is a danger that the industry will mistake managing the consequences for addressing the underlying problem. Platforms can introduce labels, detection tools and upload restrictions, but none of those measures answers the question of whether artists gave permission for their work to be used.

The problem isn't that AI can make music. It's who benefits.

There is a tendency to frame this debate as a battle between technology and tradition, as though musicians who object to AI are simply resisting change. That is not a fair representation of what is happening.

Musicians have adapted to new technology throughout the history of recorded music. They have embraced synthesizers, drum machines, digital recording, home studios and entirely new ways of distributing their work. The difference now is that the technology can be developed using the very work of those musicians, without necessarily involving them in the process or sharing the commercial benefits.

That is where the industry should be asking much harder questions.

It is not enough to tell artists that AI is inevitable, that they need to adapt or that the finished music will simply be labelled. Transparency matters, but it doesn't give someone permission to use another person's work. Nor does it compensate the people whose recordings, compositions and performances may have helped train the technology.

There is also something deeply concerning about the idea that musicians must now compete with systems capable of generating music at scale, potentially using the characteristics of the artists they are competing against. If the response is simply that artists should use AI too, we risk making the technology's commercial interests the priority and the musician's livelihood an afterthought.

Nobody is arguing that every use of AI in music should be prohibited. Artists may choose to use it in their own creative process, and that decision should be theirs. But there is a significant difference between choosing to use a tool and having your work used to build one without your permission.

Music is not just a product to be generated as cheaply and quickly as possible. Its value comes from the people who write it, perform it, take risks with it and give it meaning. An industry that forgets that is in danger of undermining the very thing it exists to support.

Musicians have already raised the alarm

Concerns about the use of creative work to train AI systems are not new, and some of music's most established names have made their opposition clear.

In 2024, Thom Yorke and other members of Radiohead, Robert Smith of The Cure and ABBA's Björn Ulvaeus were among the musicians who supported an open letter warning against the unlicensed use of creative work by AI companies.

Nick Cave has also been openly critical of AI-generated lyrics written in his style, describing the results as a grotesque mockery of what it means to be human.

That concern was also at the heart of Is This What We Want?, a protest album released in February 2025 featuring more than 1,000 musicians. Rather than containing conventional songs, the record consists of recordings of empty studios and performance spaces.

It was created in response to proposals by the UK government that would have allowed AI developers to use copyrighted material to train their systems unless rights holders opted out.

The album included contributions and support from artists including Kate Bush, Annie Lennox, Damon Albarn, Imogen Heap and Hans Zimmer.

The message was deliberately stark. If musicians lose control over how their work is used, the consequences could extend far beyond individual careers.

Who owns the music AI learns from?

This is where the debate becomes more than a question of whether AI-generated music is good or bad.

The central issues are consent, copyright, licensing, attribution and payment.

If a technology company wants to use copyrighted recordings or compositions to train a commercial AI system, should it need permission? Should musicians be paid? Should they be able to refuse? And how can they establish whether their work has been used in the first place?

These are not straightforward questions, particularly when rights are split between performers, songwriters, publishers and record labels.

The House of Lords Communications and Digital Committee addressed these issues in its March 2026 report, AI, Copyright and the Creative Industries. It warned that creative works were being used to train AI systems without explicit consent or remuneration, while limited transparency made it difficult for rights holders to establish whether their work had been included.

The committee recommended that the government rule out a new commercial text and data mining exception based on an opt-out model, and instead support a licensing-based approach to AI development.

It also highlighted the need for stronger protection against unauthorised digital replicas and greater transparency around AI training data.

For musicians, these decisions could have lasting consequences. Without clear rules, the commercial value of music may increasingly be captured by companies building and operating AI systems, rather than the people whose work helped make those systems possible.

And this is not a problem that can be solved by asking individual musicians to protect themselves. Most independent artists do not have the resources to investigate how their work has been used, challenge a technology company or negotiate complex licensing arrangements. The burden should not fall on the individual creator to prove that their rights have been ignored.

The future of music shouldn't be decided without musicians

At Beyond The Music in October 2026, industry leaders, artists and organisations came together to discuss the future of music and artificial intelligence. The discussions formed part of the Creative Industries AI Summit, where The Manchester Principles were presented.

The principles call for greater protection of creative work, fair payment where AI systems are built using it, and a stronger role for creators in decisions about how their work is used.

Among their central ideas are that creators should have control over their work, voices and likenesses, and that if AI is built on creative work, the people responsible should receive fair payment.

They are intended as a shared framework rather than a set of enforceable rules. Their significance will ultimately depend on whether those principles translate into meaningful action.

Musicians need to be part of that process. Not simply invited to comment once the commercial decisions have been made, but involved in shaping the rules around licensing, consent, attribution and payment.

AI is not going away. But accepting that reality does not mean accepting every use of it, or allowing musicians to become an afterthought in decisions about the future of their own industry.

Artists should not have to surrender their rights, identities or livelihoods to accommodate a technology they did not ask for. Nor should they be forced to adopt it simply to remain competitive.

There is a difference between keeping up with change and being forced to accommodate it on someone else's terms.

The music industry has a choice to make. It can treat AI as another commercial opportunity and leave musicians to negotiate the consequences individually, or it can establish proper protections before the imbalance becomes even harder to correct.

The responsibility cannot sit solely with individual artists trying to protect their work. Technology companies need to be transparent about what they use to train their systems. Platforms need meaningful policies. Governments need to establish clear rights around consent, licensing and payment. And musicians need a genuine voice in decisions that will shape their profession.

AI-generated music may have a place in the future of the industry. But that future should not be built on the assumption that the people who made music worth learning from can simply be left out of the equation.

The question isn't whether AI can make music. It's whether the music industry still values the people who do.