The backlash against AI from "creatives" is often presented as a single, unified protest. Artists, musicians, and authors are grouped together as though they face the same threat, for the same reasons, from the same technologies.

They don't.

Treating these groups as interchangeable obscures the real issues around copyright, learning, and commercial use — and, in doing so, allows legitimate concerns to become entangled with something else entirely: the defence of scarcity as status.

This distinction matters. Some objections to AI are serious and unresolved; others collapse under even light scrutiny. And in those cases, what remains is not a defence of creativity, but a form of elitism that sits uncomfortably with how many artists and musicians still describe themselves: anti-establishment, anti-gatekeeping, anti-hierarchy.

Authors are different — and their case is stronger

Authors occupy a genuinely different position in the AI debate, and it's worth being clear about why.

Large language models operate directly in the author's medium. Text in, text out. They can reproduce not just ideas, but sentence structure, cadence, and tone — and in some cases, passages that sit uncomfortably close to verbatim reproduction. Attribution, originality, and voice are central to literary value, and LLMs put real pressure on those concepts in ways that visual and musical models do not.

This is where questions about training data, consent, and copyright deserve serious attention. If a system can convincingly substitute for a living author in the same medium, the legal and ethical frameworks governing intellectual property begin to creak.

As a recent example, in 2024 a US federal judge ruled that using copyrighted books to train an AI model did not, in itself, violate copyright law. The case, brought by several authors against AI firm Anthropic, failed on the grounds that the training process was "exceedingly transformative" and therefore permissible under US law. The ruling did not suggest that AI systems are free to reproduce protected works or impersonate living authors; rather, it reinforced a narrower principle — that learning from material is not the same thing as copying it. More cases are being litigated, and the legal landscape will continue to shift — but the direction of travel seems clear.

In short: authors are not wrong to feel uniquely exposed. Their objections are specific, technically grounded, and unresolved. The problem begins when those objections are treated as representative of all creative work.

Learning has never required consent

One of the most common claims made against AI is that it has been trained on copyrighted material without permission. At first glance, this sounds damning. Look more closely, and it becomes incoherent.

All artists learn by absorbing copyrighted work. Always have.

Painters study the masters. Musicians learn by playing other people's songs. Writers internalise structure, rhythm, and voice by reading obsessively. Go into any major gallery and you'll often find art students copying canonical works brushstroke by brushstroke — explicitly as part of their training.

None of this has ever required consent from the original creator. And no serious creative culture has ever suggested that it should.

Much of the confusion here comes from a failure to distinguish between training and inference. Training is the process by which a model absorbs statistical patterns from large volumes of data. Inference is the moment a user asks the model to produce something new. Copyright law has always cared far more about outputs than internal learning processes, and for good reason. Training determines capability; inference determines whether that capability is abused.

The ethical boundary has never been learning. It has always been output:

  • You may study copyrighted work.
  • You may not reproduce it and claim authorship.
  • You may not misrepresent its provenance.

That distinction already exists. Extending it to machines is not radical; it is consistent.

To argue that AI training itself is illegitimate is to argue that learning must be permissioned — a position that would invalidate how every artist, musician, and writer alive today acquired their skills.

Style is not property — identity is

Where the objections do sharpen is around impersonation.

Allowing systems to generate work "in the style of" a named, living artist crosses a line. Not because style itself is ownable — it never has been — but because naming an artist collapses abstraction into identity. It invites confusion, trades on reputation, and creates market substitutes that lean directly on someone else's brand.

This is not about protecting feelings. It is about preventing misdirection.

UK law already contains a concept that maps cleanly onto this boundary. The common law doctrine of passing off protects goodwill and reputation from misrepresentation, even where no copyright or trademark applies. What it does not protect is style in the abstract. To succeed, a claimant must show goodwill, misrepresentation, and damage — in other words, that someone has falsely suggested a connection or endorsement and caused harm as a result. This is precisely the problem with AI systems that allow outputs to be framed as if they were associated with a named, living artist. The issue is not influence; it is confusion.

Platforms that block artist-name prompts are not stifling creativity; they are enforcing a sensible boundary. You can still ask for a genre, a mood, a technique, a tradition. What you cannot do is outsource someone else's identity.

That line is clear, enforceable, and worth defending. It is also far narrower than the blanket objections often made in public discourse.

The Christmas card problem

Much of the outrage directed at AI collapses when confronted with an inconvenient question: what was the alternative?

If I use AI to create an original Christmas card for my family, I haven't displaced an artist. I was never going to commission one. If I generate a short jingle for a small business experiment, I haven't put a composer out of work. The alternative wasn't paid creative labour; it was no output at all.

AI enables expression in cases where expression previously wouldn't have happened. That is expansion, not substitution.

This matters because many objections assume displacement where none exists. They treat all creative output as if it were competing in the same commercial market, when in reality most AI-assisted creativity lives well below the threshold where professional artists ever operated.

Calling that theft isn't protection. It's a category error.

Flooded markets and aesthetic panic

Another frequent complaint is that AI will flood creative markets with low-effort content. This is undoubtedly true — and largely irrelevant.

Markets have always been flooded with low-effort content: self-published books, stock photography, royalty-free music, templates, presets, loops. These things exist because they meet demand at a price point where bespoke work never would.

AI does not change that dynamic; it accelerates it.

If AI-generated content is genuinely poor, it will be ignored. If it is "good enough," then it is replacing something that was already commoditised. And if it sells, then it is meeting a market need that higher-end creative work was never serving.

There is no moral right to scarcity of output, nor to a quiet marketplace.

Power asymmetry isn't new — it's just moved

Concerns about power imbalance between creators and large AI companies sound structural, but they often rest on selective memory.

Creative industries have always been asymmetrical. Record labels, publishers, studios, platforms, galleries — intermediaries have long captured disproportionate value. AI does not introduce that imbalance; it exposes how fragile most creative income already was.

At the top end, the system has been extraordinarily generous. A single hit song can generate royalties for life. Intellectual property law has rewarded outsized success lavishly. Complaints about asymmetry often ring hollow when they come from sectors that already benefit from extreme winner-takes-most dynamics.

What's happening now is not exploitation so much as transition: a shift in where value accrues, and on what terms.

From anti-establishment to gatekeeping

This is where the charge of elitism becomes unavoidable.

For decades, many artists and musicians positioned themselves as anti-establishment — resisting labels, institutions, and corporate control. But when the tools of expression themselves become accessible, the rhetoric often changes. The fight is no longer about who controls distribution, but about who is allowed to create.

Arguments that creativity must be earned through formal technique, prolonged suffering, or sanctioned pathways are not anti-establishment. They are exclusionary. They defend hierarchy under the language of authenticity.

By elitism, I don't mean wealth or fame. I mean the belief that expressive capability should remain scarce — and that tools which lower that barrier are inherently illegitimate.

That position does not protect art. It protects status.

A narrower, stronger debate

AI does raise serious questions about authorship, attribution, and impersonation. Authors, in particular, face challenges that deserve careful legal and cultural thought. But lumping all creative objections together muddies the water and inflates claims that don't survive inspection.

AI does not abolish creativity. It abolishes the assumption that expressive power must be scarce.

For some creators, that is a genuine threat. For others, it is simply the end of a gatekeeping role they mistook for a moral one.

And that — not the technology itself — is where much of the outrage really lives.