The unease settling over museums
You are standing in front of a painting. A night scene, deep colours, a mysterious atmosphere. You look at it for a long time. Something moves you, though you are not quite sure what. You read the label: "AI-generated work, 2025". How do you feel?
Most people who have been put through the experiment describe a strange sensation. The painting does not change. But the relationship to the painting changes radically. Part of the emotion evaporates. Or rather, it turns into something else: a technical curiosity, a distance, sometimes a slight disappointment, sometimes a different kind of fascination.
Why? The painting is exactly the same. The colours are identical. So is the composition. If it was beautiful 30 seconds ago, it should still be beautiful. Yet the information about its origin profoundly alters the aesthetic experience. That alteration reveals something important about what art is to us, and about what we are looking for when we look at a work.
This article is a reflection on that question. Not a denunciation of AI art. Not a glorification either. Just an honest attempt to understand why this technology confronts us with questions we had long avoided asking.
What we asked of art before AI
To understand the current unease, we need to understand what we have historically sought in art. For most of human history, art served several very specific functions.
A decorative function. Embellishing objects, buildings, bodies. A carved door, a Gothic cathedral, a Maori tattoo. Art as an experience of beauty, capable of transforming a space or an object into something that elevates.
A religious or ritual function. Representing the sacred, telling myths, celebrating ancestors. The cave paintings of Lascaux (a prehistoric cave in Périgord, France, whose paintings date back around 17,000 years), Byzantine frescoes, African masks, medieval altarpieces. Art as an intermediary between humans and something that transcends them.
A political function. Glorifying the powerful, telling the story of a people, fixing a collective memory. Equestrian statues in public squares, history paintings in the Louvre, the propaganda poster, the Statue of Liberty.
An emotional and expressive function. This is the function that modernity gradually isolated and glorified. The artist expresses what he feels, shares a singular vision, conveys what no speech can say. Vincent van Gogh painting his sunflowers in an asylum room. Pablo Picasso and Guernica. Edvard Munch and The Scream.
A function of bearing witness. Documenting an era, a place, an experience. Vermeer's paintings (a seventeenth-century Dutch painter, master of intimate light) of Dutch domestic life. Robert Capa's photographs during the Spanish Civil War. Bob Dylan's songs about the American civil rights movement. Art as the living memory of what has been.
These functions are not equivalent. Depending on which one you privilege, AI poses very different problems.
Several recent studies have shown something troubling. When art lovers are shown two similar paintings (one human, one AI) without being told their origin, they rate the AI slightly higher on technical criteria (composition, colour, harmony). But as soon as they are told which work is human and which is AI, their preferences reverse: they overwhelmingly prefer the human work. This reversal is not a rationalisation: it reveals something profound about what we are really looking for in art.
What AI does very well, and what it does not do
To analyse the question honestly, we need to distinguish what AI really does well from what it does not do.
What AI does excellently. Pure technical mastery. Reproducing a style, harmonising colours, composing a frame, writing a grammatically perfect text. On every measurable parameter of a work, AI reaches and often exceeds the human professional level. As we documented in our article on Hell Grind, the first entirely AI film screened at Cannes in 2026, these capabilities are now accessible to any creator.
What AI does moderately well. Stylistic innovation. AI is trained on existing art. It recombines, hybridises, optimises. But inventing a completely new visual grammar (what Picasso, Pollock and Rothko did in their time) remains beyond its reach. It can produce images "in the manner of", not images that open up new territory.
What AI does not do at all. Bear witness to a lived experience. When a human paints their pain, their fears, their joys, there is a life behind the painting that has been lived. AI has not lived. It simulates, but does not bear witness. It produces, but does not put itself on the line. This is probably the difference that creates the unease in museums.
The "skin in the game" criterion
To understand what we are really looking for in art, we need to introduce a concept popularised by Nassim Nicholas Taleb (American-Lebanese philosopher and statistician, author of The Black Swan): "skin in the game". In French, it is hard to translate: having your skin in the game, taking a personal risk in what you do.
When Van Gogh paints, he has his skin in the game. His mental health, his poverty, his despair: it is all in the painting. When Beethoven composes the Ninth Symphony, he has gone deaf. His deafness, his struggle, his will to pass something on in spite of everything are in the music. When a poet writes, he reveals a part of himself that he could not express any other way.
AI has no skin. No life to lose. No suffering that expresses itself. No joy that overflows. It produces the way a factory produces: with no intention of its own, no personal commitment, no risk. And that is probably what art lovers feel intuitively when they are told that a work is AI. Not that it is less beautiful. But that there is no one behind it.
That intuition is not naive. It reveals what art has really been for 30,000 years: a mode of communication between human consciousnesses. The artist transmits something to a viewer, a listener, a reader. That something passes through the work, but the work is only a vehicle. Without a conscious sender, the vehicle still exists, but the message changes in nature.
The fields where AI is already everywhere (and where nobody has a problem with it)
The argument needs nuancing, because in several fields, AI art is already massively present, and bothers almost nobody.
Commercial illustration. Advertising posters, video game covers, illustrations for blog posts, visuals for social media. In all these uses, AI now produces a major share of the content. Nobody expects "skin in the game" in a photo of a car for an advert. The content is functional, not expressive.
Ambient music. While you read what follows, you may be listening to music. If it is generic music (chillout, lofi for studying, ambient jazz), there is a good chance it was generated by AI. Suno, Udio, MusicGen produce hours of ambient music that Spotify and YouTube distribute massively. Nobody is mourning that transformation.
Illustrations for educational content. Diagrams to explain a concept, infographics to summarise data, illustrations for children's books: AI produces perfectly satisfactory content, sometimes better than what poorly paid illustrators would have produced. The problem here is more economic (job losses for illustrators) than aesthetic.
Personalised marketing content. When Netflix generates a personalised poster thumbnail for each user, AI does the work. Nobody feels dispossessed.
In all these cases, what is asked of the image or the sound is functional, not existential. We want something that decorates, that informs, that sets a mood. We are not seeking communion with a human consciousness.
Art is gradually separating into two distinct categories: the functional (where AI will do most of the work) and the expressive (where human value remains central). This is not the end of human art, it is the end of a certain confusion. For two centuries, the same artists were asked to fulfil both functions (professional illustrator and visionary creator). The split that is taking hold may be healthier for genuine creators.
What could paradoxically strengthen human art
Here is the unexpected angle. The emergence of AI in art could, paradoxically, strengthen the value of art that is properly human. There are several reasons for this.
First movement: distinctive rarity. If anyone can generate an image in 3 seconds, the generated image loses its economic and symbolic value. But the image made by hand, with time, with effort, with commitment, takes on a relatively higher value. This is exactly what happened with photography in the 19th century. When photography became easy, painting paradoxically acquired a higher artistic status than it had ever had.
Second movement: the return to the real. The first signs are there. The Saatchi Gallery in London held an exhibition in 2025 called "All Human", explicitly presented as a "100% AI-free" zone. The Louvre inaugurated a new banner in 2026, "Touch of the Hand", for its contemporary exhibitions, highlighting works made by hand. The original art market is beginning to polarise: on one side, NFTs and AI generations, which are collapsing in value; on the other, unique physical works, which are rising.
Third movement: live performance. Recorded music is technically perfect. But live concerts have never been so popular. Beyoncé, Taylor Swift, Coldplay: their 2024-2026 tours are the most profitable in history. Why? Because a concert offers something no recording can give: shared presence, the moment that does not repeat itself, the collective emotion of thousands of people. At a time when everything becomes reproducible by AI, the irreducible (presence) takes on an insane value.
Fourth movement: signature and traceability. The C2PA standard (Coalition for Content Provenance and Authenticity) now makes it possible to cryptographically certify the origin of a work. As we explored in our article on deepfakes, this technology could reverse the problem: proving that a work is human rather than proving that it is AI. Works with certified human authentication could become a distinct category of the art market.
The question no artist can avoid
For an artist creating today, the emergence of AI poses a hard but inevitable question. Why am I doing this?
If the answer is "because I make a living from my art by fulfilling commissions", then AI is probably an existential threat. Competition at 1/100th of the price will arrive very quickly. This is what commercial illustrators, agency graphic designers and stock music composers are experiencing. Part of their trade is disappearing, and it is hard.
If the answer is "because I have something to say that no one else could say in my place", then AI is not a threat, it is almost an invigorating challenge. How can you be more accurate, more profound, more authentic in a world where the easy image has become cheap? How do you manufacture meaning in a place where the machine cannot go?
The artists who will probably survive over the next 20 years are those who will have found a profound answer to this second question. Those who have not will have to retrain or accept growing precarity. It is hard, but it is the reality.
Art as a reminder of humanity
Beyond the artistic debate, this transformation can be seen as a philosophical opportunity. By doing everything, AI forces us to ask what is specifically human. And art is probably one of the best angles from which to answer that.
When a human paints, writes or composes, they do something that goes beyond the production of content. They say "I was here, I saw this, I felt this, and I wanted you to feel it too". This communication between consciousnesses is not a luxury or a cultural accessory. It is probably one of the things that fundamentally distinguish humans from other animals and, now, from machines.
Marc Quinn, the British sculptor known for making a self-portrait with 4.5 litres of his own frozen blood, said something that sums the matter up well: "Art is not a decoration. It is proof that we were here."
AI images, however beautiful they may be, are not proof. They bear witness to no one. They remember nothing. They do not die with an artist who burned themselves out to produce them. It is this absence that creates the unease, and it is this absence that, paradoxically, highlights what we have always sought in art: the trace of a human who lived, who saw, and who wanted to pass it on.
To go further, you can read our article on Hell Grind, the first AI feature film presented at Cannes, which explores this boundary concretely in cinema, and our analysis of scientific discoveries made by AI, which raises a similar question for scientific creation. Human creation is not dead. It is in the process of rediscovering what it may always have been: an act of presence and of transmission, where the machine cannot go.