Skip to content

Notes from the Edge of Music

If AI Makes the Music, Who Is the Artist?

On authorship, synthetic voices and the difference between making music and making it matter

Amini Music · 10 October 2026 · 8 min read

This essay is not legal advice.

Two identical songs: one written over years, one generated in minutes. The listener cannot tell them apart, but the law may. What that gap reveals about authorship, identity and where creative value now sits.

Imagine two songs that sound identical. They share the same melody and the same arrangement, and the voice carries the same emotion. For the listener, the experience is exactly the same.

Their origins are not. A composer wrote the first after years of developing their craft and recorded it with musicians and a producer. The second came out of a generative system: someone wrote a description, listened to several versions, chose one and released it.

The listener cannot tell them apart, but the law might. The first composition may qualify for copyright as an original intellectual creation. Whether the second does depends on the nature and extent of the human contribution.

That paradox deserves to be taken seriously. If two pieces of music are indistinguishable to the listener, why should the process behind them decide whether they are protected? And if the music alone no longer separates the human artist from the machine, what exactly makes someone an artist?

What the law protects, and whose creativity

European copyright is built on the idea of the author's own intellectual creation. The Court of Justice of the EU has repeatedly held that protection requires original expression reflecting the author's free and creative choices, notably in Infopaq (C-5/08), Painer (C-145/10), Cofemel (C-683/17) and Brompton Bicycle (C-833/18).

The principle is simple. Copyright protects human creative expression, not the mere existence of an attractive or valuable result. A composition is original because someone made creative choices about melody, harmony, rhythm and structure. A lyric is protected because its author selected and arranged words in an original way.

Now suppose a user asks for an emotional alternative rock song with a powerful chorus and an uplifting ending. The user set the direction; the system chose the melody, chords, rhythm and arrangement.

Under the prevailing European approach, whether the user can claim copyright in that composition is uncertain. A prompt alone does not necessarily establish authorship. Hours spent generating and rejecting versions do not, by themselves, turn expression determined by the system into the user's work.

The picture changes as the human contribution grows. A songwriter supplies original lyrics. A composer brings an original melody. A producer builds a distinctive arrangement or substantially reworks generated material. Those human-authored elements can remain protected even when other parts of the production were generated. A modern recording is rarely all human or all machine. It is layered, and the legal question is which layers amount to authorship.

Does copyright measure creativity, or a particular kind of creative control?

Creativity was never a solo act

We tend to picture creation as a direct line between one person and one work, as when a composer hears a melody and writes it down. The reality has always been more crowded. A film director does not operate every camera, and an architect does not lay the bricks. A record producer can shape an album without writing a melody or playing a note. Much creative work consists of directing, selecting, combining and organising.

Generative AI adds a new kind of participant. Unlike an instrument, it makes detailed creative decisions its user did not specify. Unlike a human collaborator, it is not recognised as an author. A person can conceive a project, set its direction, evaluate hundreds of alternatives, decide its final form and bring it to an audience, and still find that some of the resulting musical expression sits outside their copyright.

Legally, that is coherent. Philosophically, it exposes two ideas of creativity: producing something, and deciding what should be produced, recognising what works and building a meaningful whole. Generative AI did not invent that distinction, but it makes it impossible to ignore.

Making something versus making it matter

A curator who has never painted can still, through selection and sequence, create an exhibition that would not otherwise exist. Music has its equivalents: the producer who hears that one vocal take transforms a song, or the creative director who gives a body of recordings a coherent world.

None of this necessarily creates copyright in the underlying compositions, yet all of it can create cultural and economic value. When a machine can produce thousands of technically convincing recordings, one more recording is no longer remarkable. The hard questions move elsewhere: which recording deserves attention, why it should exist, whom it is for and how it fits a wider artistic identity.

Can an artist exist without a human performer?

Picture a fictional band with a name, a recognisable sound, a coherent visual world and a growing catalogue. Its songs return to the same themes, and listeners know what to expect from the next release. It has no conventional members. Its vocals are synthetic, and its identity comes from human creative direction combined with technological production.

Legally, the band is not a person and holds no rights. Its music, recordings, name and artwork may nonetheless involve rights that belong to identifiable people or companies. Culturally, things are less tidy. Listeners do not experience music by first working out who legally owns it. They experience songs, images and stories, and they form attachments to characters and personas.

None of this is new. Virtual bands have shown that a fictional group can become a global project while keeping its characters distinct from the musicians behind them, and every band's public identity is partly constructed. An artist has always been both a person producing sound and an identity through which an audience hears it. Generative tools extend that possibility, and raise a question: when the performer is synthetic, what is the audience connecting with?

Voice, resemblance and impersonation

The voice is not just another instrument. It carries personality and presence; we recognise singers within a few words. A synthetic voice can sound convincingly human without belonging to anyone, and occasionally it may resemble an existing singer.

That resemblance needs careful thought. Copyright and neighbouring rights protect specific expression and recorded performances. They do not generally give anyone a monopoly over a type of voice, so a rough baritone or a breathy delivery is not the property of one singer. Deliberately generating a recognisable digital imitation of an existing performer is a different matter. It can engage personality rights, rules on misrepresentation and protection against the unauthorised exploitation of someone's identity.

Article 50 of the EU AI Act adds transparency obligations for certain AI-generated or manipulated content, including deepfakes. For evidently artistic or fictional works, it allows that disclosure to be made in a way that does not spoil the work. The law on identifiable synthetic voices is still developing, but in our view one principle already stands out: creating a fictional voice is not the same as appropriating a real person's voice. Fictional artistry is legitimate. Deceptive impersonation is not.

Inside the experiment

At Amini Music we explore these questions through practical work in Amini Labs. Neon Shelter is an attempt to build a coherent fictional artist identity, with a recognisable sound, recurring themes, a visual language and a catalogue that develops over time. Releases such as Rebellious Perfume, Bad Advice and Fame Looks Ugly Up Close each test a different side of that identity. The challenge is not to make songs that sound professionally produced. It is to decide what makes them belong together, and how the sound can evolve without losing its character.

StrategyX asks the same questions in electronic music, where the work lies in the relationship between sound, image and concept. The multilingual projects develop a single song concept across languages and cultural contexts. They ask whether a song stays the same work when its language and setting change, and whether an artistic concept can travel without becoming generic.

The tools let us investigate these questions faster than we otherwise could; they do not answer them. We are not trying to prove that AI can generate music. We want to understand where human judgment remains decisive when production becomes a tool for rapid experimentation. The experiments are documented in Amini Labs.

Effort, or the lack of it, is not the measure here. Multitrack recording, sampling and pitch correction each reduced the labour behind a result and each set off a debate about authenticity; generative AI belongs to that longer history. The real distinction is different: a work can be valuable without being copyrightable, and copyrightable without being valuable to anyone.

That gap matters commercially. A music project can build a recognisable identity and a loyal audience while its rights are spread across different layers. Its artwork may be protected by copyright and its name by trade mark law. Human-written lyrics may be protected as literary works, and recordings may attract neighbouring rights. Generated musical elements have a protected status that depends on the human contribution involved and remains legally unsettled.

That is not a contradiction. Businesses have always created value from assets that are not fully owned in the intellectual property sense, such as reputation, know-how, market access and relationships. Generative music makes this layering more visible and raises practical questions. How should such a venture be valued? Which assets can be licensed exclusively? How should human contributions be documented? Can an artistic identity survive a change of production method? The legal architecture of a music venture may become as important as its production technology.

Authenticity is not the same as origin

It is tempting to sort music into human and therefore authentic, AI and therefore artificial. Yet a human performer can deliver an empty performance, and a synthetic recording can move people. The biological origin of a voice does not settle whether it is authentic. That depends on the relationship between expression, presentation and what the audience is led to expect.

None of this is a licence to deceive. Leading listeners to believe that a particular person performed a recording when they did not is a serious ethical and legal problem. Transparency matters, but it does not require abandoning fiction. A fictional artist can be presented as fictional, and an AI-assisted project can acknowledge its methods without turning every song into a technology demonstration. The line runs between artistic illusion and factual deception. Illusion builds a world for the audience to enter. Deception misleads the audience about the world outside the work.

A better question

The usual question is whether AI can create art. The answer depends so heavily on definitions that the question rarely gets us far. A more useful one is this: which forms of human creativity become more important when machines can generate convincing material? They include recognising potential, developing concepts, making aesthetic judgments, understanding audiences, building coherent identities and knowing when to stop generating and start building.

These skills always mattered; the difficulty of production often hid them. They may not create copyright in every element, but they decide whether a project means something or disappears into an endless supply of interchangeable recordings.

Producing music and being an artist are becoming different activities. Law, technology and markets will each shape what follows. None of them settles what makes music meaningful.

A machine can generate a song. Someone still has to decide why that song should exist.

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

Sources and further reading

Part of Notes from the Edge of Music, Amini Music's series on music, technology, rights and new creative business models. Read how we work with AI in our AI, Rights & Creative Production Policy.

Tags: Authorship, Artist identity, Synthetic voices, Copyright, Creative direction