Who Owns the Future of Music?

For most of recorded music's history, a commercially viable song needed scarce resources: musicians, studios, producers, equipment, distribution and, usually, a label's attention. Technology removed those barriers one by one. Digital recording made professional production affordable, the internet broke distribution monopolies, and streaming put an almost unlimited catalogue in every pocket.

Generative AI challenges something more basic: the scarcity of production itself. A song that once took weeks of collaboration can become a convincing recording in minutes, and a fictional artist can acquire a voice, a catalogue and a visual identity without entering a studio.

That opens real opportunities for independent creators. It also raises uncomfortable questions for an industry built on owning and exploiting intellectual property. Who owns the resulting music? Who should be paid when existing works are used to train these systems? Who is responsible when a model produces something that resembles an existing composition?

These questions are now in court. The dispute between the German collecting society GEMA and the generative music platform Suno shows that a larger question lies beneath them: how economic power in music will be distributed.

What the Munich court decided

On 31 July 2026 the Regional Court of Munich I ruled in proceedings brought by GEMA against Suno. The case concerned the reproduction of protected musical material by Suno's system, and the works examined reportedly included Forever Young and Daddy Cool. According to the court's press release and published commentary, the court found that the system could reproduce protected musical elements. It held the provider responsible for the infringements established in the proceedings.

The decision is significant, but it is not the final word. It is a first-instance German judgment, and its reach depends on the court's reasoning, the facts before it and any further proceedings.

It does sharpen a basic problem. Generative systems learn patterns of melody, harmony, rhythm, instrumentation, arrangement and vocal character, and much of that comes from existing music. Existing music is not just data, though. It is human creative work, often protected by copyright and neighbouring rights. When a company uses it to build a system that produces commercially exploitable recordings, is the company learning from culture or exploiting protected works?

The answer differs from case to case. EU copyright law allows text and data mining under conditions, including the rightsholder's ability to reserve their rights, while keeping exclusive rights over reproduction. The EU AI Act adds that providers of general-purpose AI models must maintain a copyright compliance policy and publish a summary of the content used for training. Litigation is now testing where those frameworks actually draw their lines.

GEMA's concern is understandable: creators should not lose control over the exploitation of their work just because it happens through a new process. The developers' concern is real too. If training requires licences across vast catalogues, building competitive models becomes far more costly. Both interests are legitimate, and neither tells the whole story.

Learning is not copying

Human creativity has always depended on earlier work. Composers study composers, producers absorb whole movements, and genres grow through constant exchange. So why treat an AI system differently when it learns from existing music?

The analogy is attractive but incomplete. A model is trained through technical reproductions and at a scale unlike any musician's learning, and its output can be distributed at a scale no musician could match. More importantly, copyright does not prohibit influence. It regulates specific acts involving protected expression, and a style is generally not protected in the way an original melody is.

A request for an energetic rock track with distorted guitars and a rough male vocal therefore does not infringe anyone's rights simply because the result follows the conventions of the genre. If the system reproduces a recognisable protected melody, the analysis changes. The real boundary does not run between human and machine creativity. It runs between the legitimate use of musical knowledge and the unauthorised reproduction of protected expression.

That leads to the question of responsibility. If a model produces infringing material in response to general instructions, who is responsible? It could be the developer who trained the model, the user who asked for a new song, the distributor who released it, or some combination of them. A user who deliberately instructs a system to reproduce a known song is in a different position from someone who asks for an original composition and receives an unexpectedly similar result. A framework that treats the two alike creates uncertainty far beyond the conduct it aims to stop.

Copyright is also about market power

It is tempting to read the case as creators who deserve payment against technology companies trying to avoid it. There is truth in that. Creators invest skill and time, and their rights should not dissolve because a new intermediary finds a way to extract value from their work.

Copyright does more than protect expression, though. It creates exclusive rights that can be bought, licensed and concentrated, and in practice the power attached to those rights is unevenly spread across creators, publishers, labels, collecting societies and technology firms. If access to training material depends on expensive licences, large technology companies will negotiate them more easily than small ones. Major rights owners will also capture more of the revenue than independent creators. A licensing regime could compensate existing owners while making it harder for new developers to enter the market.

That does not make licensing wrong. It means that the structure of licensing matters as much as the principle of compensation. Who receives the money, and how is it distributed? Can independent developers obtain licences on reasonable terms? Are individual creators properly represented? Protecting creators and protecting existing business models are not always the same objective.

When production is no longer scarce

Generative AI pushes the cost of a plausible recording towards zero. That does not make every generated song good, but it does move value elsewhere.

Photography offers a precedent. Digital cameras, and then smartphones, made taking pictures almost free, yet professional photography survived. Its value moved to direction, context, quality, reputation and access. Music may follow the same path. If almost anyone can produce a competent recording, recordings become less scarce. Attention, trust, a recognisable identity and an audience that cares remain scarce.

The ability to produce music may become a commodity. The ability to create meaning, identity and demand around music may not.

For established rights owners, this means catalogues competing with a near-unlimited supply of new recordings. For independent creators, it means a strong idea no longer needs a label or a large budget to find out whether an audience exists. It also blurs familiar roles. Someone may write original lyrics, set the musical direction, generate and reject arrangements, edit the strongest version, build an artist identity and release the result without playing or singing a note. The categories of musician, producer, creative director and entrepreneur are shifting.

None of this means that prompting creates copyright, because EU law still requires identifiable human creative expression. Copyright is only one dimension of value, however. A coherent catalogue, a recognisable identity and an audience can be valuable in their own right. That is why some future music businesses may look more like creative venture studios than traditional labels.

The missing party: the independent creator

Public debate usually sets established rights owners against AI companies. There is a third group: independent creators and producers who use generative tools without controlling them. They have no say over training data, model architecture or design choices, and their position should not automatically be equated with that of the company that trained the model.

Users cannot expect unlimited immunity either. Deliberately reproducing a protected composition, uploading recordings without permission or creating a convincing imitation of an identifiable performer can all have legal consequences. What is needed is a framework that separates intentional infringement from legitimate experimentation and unforeseen technical outcomes.

The Munich ruling helps here. The court examined outputs that reproduced protected expression even though the prompts did not prescribe those musical details, and it placed responsibility for the infringements it established on the provider (Regional Court of Munich I). That does not give users blanket immunity, but it points to the right question: who actually controlled the infringing reproduction? If the law ignores that question, independent producers face risks they cannot investigate or control. Only organisations with large legal budgets would then be able to experiment, which would be a strange outcome for a technology whose promise is wider participation.

Licensing is necessary, not sufficient

Litigation is likely to push rights owners and AI developers towards broader licensing, and the market is already moving that way. In September 2026 Suno announced a licensing partnership with Believe, according to Music Business Worldwide. Litigation and commercial cooperation are developing side by side.

Licensing will not settle everything, however. It will not decide whether a particular generated melody infringes an existing one, or what the copyright status of predominantly generated music is. Nor will it settle how synthetic voices and fictional performers should be treated, or how value is divided among those who create, finance and distribute music. If licensed tools become so expensive that only established companies can use them, the market loses much of what generative technology could offer. The aim should not be to make AI music expensive. It should be to let innovation and fair compensation coexist through careful legal distinctions, transparent licensing and real access for smaller players.

Who captures the value

The history of the music business is partly a history of shifting control over scarce resources. Labels once controlled recording and distribution, and streaming platforms later came to control access to listeners. Generative AI companies may now shape how music is created as well as how it is distributed.

AI will not make copyright less important. If anything, it makes ownership, licensing and provenance more consequential. Copyright alone will not decide which artists find audiences or which business models emerge, though. Creators deserve meaningful protection and payment, and developers need predictable rules. Independent producers need access to tools without unlimited liability for processes they do not control, and audiences benefit from a music ecosystem that stays diverse and open.

The GEMA–Suno case is an early contest over the conditions under which a new production infrastructure may operate. The challenge is to protect creative expression without protecting scarcity for its own sake. The future of music is not a choice between human creativity and artificial intelligence. It is a question of how the two coexist, and of who gets to take part.

Why this matters for Amini Music We use AI as a production tool, so the line between general musical instruction and reproducing protected work is a working reality for us, not an abstraction. Our AI, Rights & Creative Production Policy sets out how we approach it: no intentional reproduction of third-party works or voices, care with source material, and a route for rightsholders to raise concerns.

This essay is commentary on legal and market developments as of October 2026. It is not legal advice.

Sources and further reading

• Regional Court of Munich I — press release on GEMA v. Suno, 31 July 2026
• GEMA — court decision concerning Suno
• DLA Piper — analysis of the GEMA–Suno judgment, 28 August 2026
• Directive (EU) 2019/790 — Copyright in the Digital Single Market
• Regulation (EU) 2024/1689 — Artificial Intelligence Act
• Music Business Worldwide — Suno's licensing deal with Believe

Part of Notes from the Edge of Music, Amini Music's series on music, technology, rights and new creative business models.