GP / Blog / Arts & Humanities

Issue forecast

AI and creativity: can a machine make art, and does the maker matter

The value and nature of art is one of the bank's most loyal returners. The newest form of the question is the hardest one yet: when a machine can produce a passable song, a painting or a poem on demand, is the output art, and does it matter that no one felt anything making it?

Theme · Arts & HumanitiesLast set at A-Level · a new angle on a recurring theme (the value-of-art strand was last set in 2025, on whether only well-known art can be great)Foresight, not a prediction

In short

An AI-and-creativity question rewards a student who can hold two definitions of art apart: art as an object you experience, and art as an act of human expression. Argue whether the maker matters, concede the strongest point on the other side, and use the real copyright fights and the flood of AI music rather than abstract worry about robots.

Why this could come up now

The nature and value of art is a hardy perennial that keeps changing vehicles. It ran in 2016 as adaptations against originals, and most recently in 2025 as whether only well-known art can be great. The strand is reliably present; AI is simply its newest and most pressing form.

It is impossible to miss in the news an examiner reads. In September 2025 Anthropic agreed to pay 1.5 billion US dollars to settle a copyright case brought by authors, music publishers followed with their own multi-billion-dollar suits, and by April 2026 around 44 percent of new tracks uploaded to one major streaming service were AI-generated. The question of what counts as creativity has moved from philosophy to the courts.

Framed honestly: this is foresight, not a forecast of wording. The value-of-art strand recurs, and the AI angle is the freshest way it could be set, so it is worth rehearsing even though no one can promise this is the year it lands.

201620192021202320252026

set at A-Level most recent appearance. The nature-and-value-of-art strand recurs every few years; AI is the newest vehicle for an old question.

What an essay on this would test

These questions test whether you can define your terms before you argue. 'Can AI make art' is unanswerable until you decide whether art means the finished thing that moves a listener, or the human act of making meaning. Settle that, and the essay almost writes itself; dodge it, and you talk past yourself.

They also test whether you can resist both the hype and the panic. The strong script grants what generative tools genuinely do, lowering the barrier so more people can make something, before pressing the harder claim about whether intention and experience are part of what art is.

Operative angles

art versus the artefact: a beautiful output you experience, or a human act of expression and intention, which are not the same thing
does the maker matter: whether knowing a machine made it changes what the work is, or only how we feel about it
creativity versus generation: producing a novel combination from training data, against originating meaning from lived experience

How to answer it: two ways in

Two distinct, defensible routes through the question. A strong script commits to one and uses the other as the concession it answers, rather than sitting on the fence.

The result is what counts

A tool that widens who gets to create

Art has always advanced on new tools, and if the output moves an audience, demanding a human author behind it is a romantic prejudice, not a property of the work itself.

Every medium, from the camera to the synthesiser, was once dismissed as not real art, then absorbed; generative AI is the newest tool in that line.
These tools democratise creation, letting people without years of training or money make music, images and writing they could not before.
We routinely respond to a work before knowing who made it, which suggests the experience the work creates, not the biography behind it, is where its value sits.

Worked exampleBy April 2026 around 44 percent of new tracks uploaded to the streaming service Deezer were fully AI-generated, on the order of 75,000 a day, evidence that the tools have already put creation in far more hands, whatever one thinks of the results (Deezer, as of 2026-06).

Art is human expression

Output without intention only mimics art

Art is not just a pleasing artefact but a human act of meaning-making, so a machine that recombines existing work produces something that resembles art without being it, and the maker matters because expression is the point.

A model originates nothing; it remixes what humans already made, which is why so much of the value debate is really a fight over whose work it was trained on.
Creators have judged the threat serious enough to strike and to sue, which is hard to square with treating AI output as simply more art.
If art is communication between people, an artefact that no one meant anything by can move us by accident, but it is not the same act as a person trying to say something true.

Worked exampleIn 2023 Hollywood's writers and actors struck for months and won contract terms that a writer or performer must be human, and in September 2025 Anthropic agreed to pay 1.5 billion US dollars to settle authors' claims that their books were used to train its models, both signs that creators treat AI output as built on human work, not a substitute for it (WGA and SAG-AFTRA 2023; Anthropic settlement, as of 2026-06).

The fuel: stats, facts and examples

44%

of new tracks uploaded to the streaming service Deezer were fully AI-generated by April 2026

Source: Deezer Newsroom · as of 2026-06

US$1.5bn

Anthropic's September 2025 settlement with authors over books used to train its AI, around 500,000 works covered

Source: Anthropic settlement, reported by NPR and CNBC · as of 2026-06

2023

the year Hollywood's writers and actors struck and won contract terms reserving authorship and performance to humans

Source: WGA and SAG-AFTRA 2023 agreements · as of 2026-06

US$3.1bn

lawsuit filed by major music publishers against Anthropic in January 2026 over training on copyrighted songs

Source: Universal, Concord and ABKCO suit, reported Jan 2026 · as of 2026-06

Facts worth deploying

In a 2025 ruling a US judge found that training an AI on lawfully obtained books could count as fair use, but that using pirated copies did not, which is why the legal fight is now partly about how the training data was obtained.Source: Bartz v. Anthropic, reported 2025, as of 2026-06

Deezer reported that the large majority of streams from fully AI-generated tracks in 2025 were fraudulent, used to farm royalties, which complicates any simple claim that AI is just widening genuine creativity.Source: Deezer Newsroom, as of 2026-06

The 2023 Hollywood agreements turned on a single line that a writer of literary material or the performer behind a digital replica must be human, which makes them a concrete test of whether the maker has to be a person.Source: WGA and SAG-AFTRA 2023, as of 2026-06

FAQ

Is this a technology essay or an arts essay?

An arts essay that uses technology as its case. The real question is what art is and whether the maker matters, not how the model works. Use the copyright fights and the streaming flood as evidence, then return to the definition. Every point should land on art and meaning, not on engineering.

Can I argue AI art is just theft?

You can argue the training-data point, and it is strong, but be precise. A court has accepted that training on lawfully obtained work can be fair use while pirated copies are not, so 'theft' is too blunt. The sharper claim is that AI output depends on human work it did not credit, which is an argument about originality, not a slogan.

How do I avoid sounding either hyped or alarmist?

Concede the real thing each side has. The tools genuinely widen who can create; they also genuinely unsettle creators who struck and sued. Hold both, then argue the definition. The middle band goes to scripts that only praise or only fear; the top goes to the one that decides what art is and defends it.