AI Didn't Break Art — It Exposed What We Were Really Buying

0:00The Rubens price list

Four hundred years ago, a painter in Antwerp sold the same picture at three different prices. Same canvas. Same subject. Same skill on display. What changed was the story of whose hands had touched it.

His name was Peter Paul Rubens. He ran what we would now call a studio system — assistants, apprentices, specialists who painted nothing but fabric, or animals, or sky. And in his letters, Rubens graded his own inventory with the bluntness of a man reading out a price list. "Original, entirely by my hand." "Original, with some assistance." "Copy by my pupils." The prices followed the grades.

Hold onto that, because we are about to have an argument about machines. And almost everyone having that argument believes something Rubens would have found ridiculous. The idea that we judge art by looking at it.

0:57Your eye doesn't object

Here is a question that sounds simple and isn't. Suppose I show you an image, and you find it beautiful. Then I tell you a machine made it. Did the image change?

Obviously not. The pixels are identical. Something else moved. And researchers have started measuring exactly what.

In 2025, a team at ETH Zurich published three pre-registered experiments in the journal Cognition. They showed people artwork, then varied what those people knew about how it was made. When participants were given factual information about how these systems actually generate images, their ratings fell. And not only their ratings of beauty. Their moral ratings fell too. The work became less appealing and less acceptable in the same motion.

But the part I find genuinely strange is the part most summaries skip. The same team also ran an implicit association test. That is the kind designed to catch the biases you don't know you're carrying. And down there, the effect largely dissolved. Human art and machine art did not reliably sort themselves into good and bad.

Sit with that for a second. Your eye doesn't object. Your knowledge does.

Which suggests the thing we call "ours" was never sitting inside the artwork. It was sitting in the story attached to it. And if that feels like a modern problem, I have bad news about the last five centuries.

2:29The workshop was a business

Come back to Rubens. We tell ourselves a story about art in which a solitary genius suffers alone and emerges with a masterpiece. That story is largely a marketing invention, and a surprisingly recent one. The Renaissance workshop was a business. Rubens sketched the design, painted the faces and the hands — the parts that sold — and handed the rest to men who were often extraordinary artists in their own right. One of them, Anthony van Dyck, arrived at seventeen and became so fluent in his master's style that he could stand in for him.

Three centuries later, Andy Warhol named his studio the Factory. He was not being coy. He was telling you the joke out loud.

So the boundary was never one pair of hands. It was collaborative, negotiable, and priced according to provenance rather than appearance. Rubens's clients were not buying a better-looking painting. They were buying a better story about who had made it.

Then why does this time feel different?

3:31A panic we invented

Before you answer, let me tell you about the last time we were certain art had just been murdered.

In 1839, the daguerreotype was unveiled in France, and the painter Paul Delaroche is supposed to have declared: "From today, painting is dead."

It is a perfect quote. It gets repeated constantly. It is also, as far as anyone can establish, something Delaroche never said. The line first surfaces in a survey published in 1873 — thirty-four years after the moment it supposedly captures. And Delaroche himself? He kept painting until his death in 1856. Far from recoiling, he thought well of photography.

We invented a panic, and then attributed it to a man who wasn't panicking.

What actually happened was slower, and far more interesting.

4:23Authorship lives in choices

Photography forced a legal question nobody had needed to answer before. Can a machine-made picture belong to anyone?

In 1884, the United States Supreme Court took up the case. A photographer named Napoleon Sarony had made a portrait of Oscar Wilde, and a lithography company copied it, arguing that a photograph is merely a mechanical record. The camera did the work. So where is the author?

The Court disagreed. And pay close attention to why. It did not rule that Sarony deserved protection because a human had touched the equipment. It ruled that he had posed Wilde, selected the costume, arranged the draperies and accessories, set the lighting, and chosen the expression. Authorship lived in the choices.

A human pressing a button was never the standard. A human deciding something was. Hold that, because it is about to matter enormously.

5:23The intention we abandoned

Here is where our modern argument goes quietly incoherent. The most common objection to machine-made art is that there was no intention behind it. Nobody meant anything by it. It is hollow.

That sounds reasonable. It is also a position that literary critics spent much of the twentieth century demolishing. In 1946, W. K. Wimsatt and Monroe Beardsley published an essay called "The Intentional Fallacy," arguing that the author's design is neither available nor desirable as a standard for judging a work. The poem belongs to the public. Interrogate the text, not the man.

For decades, that was sophistication. Caring what the artist meant marked you as naive.

And then a machine began making things nobody meant, and we grabbed for authorial intention like a life raft.

I don't think that makes us hypocrites. I think it makes us honest, finally. We were never as indifferent to the maker as we claimed. We simply had never been tested.

The law, meanwhile, has stopped waiting for us to sort it out. A computer scientist named Stephen Thaler tried to register a copyright for an image titled "A Recent Entrance to Paradise," naming his AI system as the author. In March 2025, a federal appeals court in Washington ruled against him. Under the Copyright Act, an author must be human. A year later, the Supreme Court declined to revisit it.

And there is a stranger precedent sitting behind that one. In 2011, a crested macaque named Naruto picked up a wildlife photographer's unattended camera in Indonesia and took a series of self-portraits. Years of litigation followed. The courts held that a monkey cannot own a copyright either.

So the line the law has drawn is not about talent, or beauty, or effort. It is a line drawn around our species. Which is honest, in its way. It is also an admission that nobody has a principled definition. It's a fence, not an argument.

7:32The training data question

Now — the part of this that isn't philosophical at all.

While we debate whether machines can have souls, a far more concrete question sits underneath, and it has nothing to do with metaphysics. These systems were trained on work. Enormous quantities of it. One widely used dataset catalogues roughly five billion images scraped from the open internet, and a great many of them were made by people who were never asked, never credited, and never paid.

In 2023, a group of artists sued Stability AI, Midjourney, and DeviantArt over precisely that. In 2024, a federal judge refused to dismiss their core copyright claims and let the case move forward. It still has not been decided on the merits. Whether training on copyrighted work counts as fair use remains genuinely open.

Let me put this plainly, because it is where the topic usually collapses into a sermon. If what makes art "ours" is provenance — the story of who made it, and what it cost them — then the training data question is not a side issue. It is the entire issue, wearing different clothes. The offense was never that a machine can paint. It is that every one of those paintings has a history, and somebody stripped it off.

8:53The case against provenance

But I owe you the strongest argument against my own position.

If provenance is what we are truly valuing, then we are one short step from pure credentialism. Judge the biography instead of the object, and you get an art world where a mediocre canvas sells for millions because of the signature on it, while a magnificent one is ignored because nobody has heard of the person who made it. That world is not hypothetical. We have built it before.

There is a second problem, and it is worse. Cost is not a virtue. A painting that took a decade is not automatically better than one made in an afternoon. Suffering is not craft. If we start treating difficulty as the measure of worth, we will end up applauding labour and calling it beauty — which is its own kind of lie.

So I will leave you with the discomfort instead of a resolution, because I don't believe the resolution has arrived yet.

For most of human history, making things was expensive. And we quietly built our entire sense of artistic value on top of that expense. That foundation is dissolving now. Not because machines learned to feel — they haven't — but because the cost that used to authenticate our work has collapsed.

Rubens had three grades for his paintings. Entirely by his hand. With some assistance. By his pupils.

Somewhere out there is a fourth line. One for the work where a person brought years of taste, a thousand rejected attempts, and one specific idea no machine would have reached for on its own.

Nobody has written that line yet.

What would you want it to say?