A $432,500 Painting: Who Is the Creator When the Algorithm Paints?

Christie's in New York sold an algorithm-generated painting for $432,500 in October 2018, about 45 times its estimate, according to Smithsonian magazine. Whom should we applaud? The French collective Obvious, who presented it, the person who wrote the code, or those who painted over the centuries what the machine learned from?

How the machine learns to paint

This art was born of an idea that appeared in 2014: generative adversarial networks (GANs), by Ian Goodfellow and colleagues. One model generates samples, another judges whether they come from real data or were generated, and each pushes the other to improve. Models then arrived that turn written descriptions into images. The difference from older attempts is that the machine no longer executes rules written by a person; it extracts aesthetic patterns from vast numbers of works.

The showing of “Belamy” was not free of dispute over credit: some criticised the collective's reliance on code published on GitHub by a young artist without sufficient acknowledgement, and one of its founders admitted they had built on his work. There is an early lesson here: creating with a machine is a chain of hands, not one hand.

Every new tool was met with fear

The camera provoked wide debate among critics and artists in the nineteenth century, and painting did not die. The Thyssen-Bornemisza Museum notes that photography pushed Manet, Degas and the Impressionists towards new ways of seeing, that the influence ran both ways, and that French critics argued over it at length. But this time there is a difference: the camera captures what the photographer places before it, whereas a generative model produces new images from the works it was trained on.

Who owns the work?

This raises two legal questions. First: does the law protect what a machine generates? The US Copyright Office concluded in its report (January 2025) that text prompts alone do not give sufficient human control over a work, and that purely machine-generated material is not protected, while original expression a human adds, or creative selection and arrangement, is. Second: is training on protected works without permission lawful? In November 2025 the High Court of England and Wales rejected Getty Images' claim that Stability AI's model was an infringing copy, since its weights do not contain copies of the works, and accepted only very limited trade mark claims. This does not settle the matter in every country; systems differ and cases continue.

Nor do I find it a sufficient answer to say that humans learn from paintings too. An artist's influence from a teacher passes through experience, context and cost, whereas a model recombines statistical patterns. Ethically, those on whose work a tool was built have a right to be consulted or treated fairly, and that is part of trustworthiness.

Where does the value go?

An experiment with 758 consultants (2023) showed that those who used AI completed tasks within its capabilities 25.1% faster and at higher quality, but were 19 percentage points less likely to reach the correct solution on a task outside its limits. The tool excels in some areas and collapses in others, and the human's value shifts from completing the work to judging: when to trust the output and when to override its average taste.

An artist can take it as a pen he bends to his vision, speeding up a draft and exploring possibilities, then supplying the intention and the choice himself. Whoever competes with it in volume and speed is unlikely to prevail; that is its arena. The tools also open a door to Arab heritage in calligraphy, ornament and architecture, to reread it and widen access to it, provided the rights of its owners and the bounds of the Sharia and public taste are respected in what is produced.

Is the machine creative?

Neither this nor that. The machine explores a space of possibilities drawn from human heritage and emerges with arrangements that may dazzle, but it intends nothing. If you see beauty in them, the meaning is one you project; a work by a person about a loss he lived carries its meaning from its maker. Perhaps the best this debate can do is return us to an old question: art is not only making beautiful things, but conferring meaning.