In January 2025, the Getty Museum in Los Angeles integrated Cristian en el Amor de Calle, an AI-generated photograph by Matías Sauter Morera, into its permanent collection. In the same year, Getty Images—a distinct entity but sharing the same name—intensified its legal battle against Stability AI in US and UK courts for massive copyright infringement. The same name. Two irreconcilable actions.
The art world has seen this scene before.
In 1859, Charles Baudelaire published his famous indictment against photography in the Revue française, calling it the "servant of the sciences and arts" and denying it any creative status. In 2023, American critic Jerry Saltz—one of the most listened-to voices in contemporary art—described Refik Anadol's installations at MoMA as a "glorified lava lamp." One hundred and sixty-four years separate these two judgments. The rhetorical structure is identical: a mechanical medium cannot produce art.
Baudelaire was wrong. Saltz might be too. Or not. The analogy deserves more than a footnote.
The Four-Act Scenario
Each time a technology penetrates the realm of visual creation, the same sequence unfolds. Initial rejection by the establishment. Adoption by pioneers who face ridicule. Gradual institutional legitimization. Coexistence with previous forms. Photography, video art, and digital art have all gone through this sequence. Generative artificial intelligence is going through it today—at a speed that historical precedents did not prepare for.
Photography, invented in 1839, took nearly a century to gain full artistic recognition. The Museum of Modern Art, founded in 1929, still gave it only a marginal place. It was not until the 1960s and 1970s that it was unanimously considered an art in its own right, with its masters, schools, and specialized criticism.
Video art compressed this cycle by half. In 1963, Nam June Paik exhibited thirteen manipulated televisions at his first solo exhibition in Wuppertal. Two years later, he filmed Pope Paul VI's visit to New York and declared the recording a work of art. Video achieved full institutional legitimacy in fifty to sixty years.
Generative artificial intelligence is accelerating the curve even further. Its dramatic entry into the art market dates back to 2018, when Edmond de Belamy, a portrait generated by a Generative Adversarial Network (GAN), sold for $432,500 at Christie's—more than forty times its high estimate. Less than five years later, in October 2023, MoMA integrated Refik Anadol's Unsupervised into its permanent collection: the first generative AI work to achieve this status. The installation had attracted over three million visitors, one of the museum's biggest attractions.
In spring 2026, DATALAND, the world's first museum dedicated to AI arts, opens in Los Angeles under the impetus of Anadol himself. The first museums dedicated to photography only appeared in the late 1940s. AI art gets its own institution one hundred and fifty years faster.
The numbers illustrate an acceleration law: one hundred years for photography, fifty for video, an estimated thirty to forty for artificial intelligence. This chronological telescoping partly explains the intensity of the reactions. What society had a century to digest with photography, it must absorb in one generation with AI.
Walter Benjamin Almost Predicted Everything
In 1936, Walter Benjamin published The Work of Art in the Age of Mechanical Reproduction, an essay in which he coined the concept of "aura." This quality of uniqueness, authenticity, and presence possessed by the original work is destroyed by mechanical reproduction. Benjamin analyzed the shift from a "cult value"—the unique, sacred work rooted in a place—to an "exhibition value"—the reproducible, circulating, democratized image.
Generative AI pushes this logic to its culmination. Reproducibility is no longer mechanical; it is algorithmic. An AI-generated image has no original in the way a silver halide photograph has a negative. It is born digital, infinitely duplicable, devoid of any aura by definition. "Exhibition value" reaches its pure form: the image exists exclusively to be seen, shared, disseminated. No material relic anchors it in the physical world.
Benjamin predicted that mechanical reproduction would democratize art while destroying authenticity. This prediction applies to AI with almost disturbing precision. The 2024 Hiscox collector survey reveals that 82% demand mandatory labeling for works produced or assisted by AI; 60% are concerned about authenticity. These collectors, without having read Benjamin, are expressing exactly the anxiety he described: faced with total reproducibility, the market desperately seeks to reconstitute aura—through labels, certification, traceability. As our survey on the social divide caused by AI in the creative world shows, this demand for transparency cuts across the entire artistic community, from collectors to illustrators.
The Pioneers and Their Historical Doubles
The role played by certain artists in legitimizing AI art faithfully mirrors that of their predecessors. Refik Anadol at MoMA in 2023 is what Nam June Paik was at the Parnass Gallery in Wuppertal in 1963: the pioneer whose practice forces institutions to redefine the boundaries of art. One manipulated televisions. The other trains algorithms on 138,151 images from MoMA's collection. The gesture is the same: to seize the dominant technology of their era and turn it into an artistic medium.
This genealogy illuminates the trajectories analyzed in our cross-portrait of artists working with artificial intelligence. Holly Herndon, a Stanford PhD and co-founder of the haveibeentrained.com platform, embodies a figure unknown in the 19th century: the artist-technologist who uses the medium and advocates for the rights of those it threatens. Sougwen Chung, a former researcher at MIT Media Lab, whose robotic arms draw from EEG readings, presented her work at the World Economic Forum in Davos in January 2025. Her approach recalls pictorialist photographers who advocated manual intervention on the mechanical process to prove their work was art. Mario Klingemann, with Botto—an AI that creates and sells its works via decentralized governance, totaling over five million dollars in sales since 2021—pushes the question of authorship into areas that even photography had not explored.
Paul Martineau, curator at the Getty Museum, justified the acquisition of Sauter Morera's work by stating that the artist uses "hundreds of commands and choices that take months." This argument is word-for-word what photography's defenders argued against 19th-century academicians: the choice of framing, light, and shutter release made the operator an artist, not a technician. The camera was merely a tool. AI is merely a tool. The pattern repeats because the underlying question has never changed: where is the creative act located?
Copyright, or History Stuttering
The legal battles surrounding generative AI reproduce the conflicts that accompanied the birth of photography. In the mid-19th century, Nadar and his colleagues had to plead in French courts for the law to recognize photography as an act of creation, not a mechanical product. Detractors' argument was one sentence long: the machine does the work, the photographer is merely an intermediary. It took decades for the law to adapt.
As our deciphering of the legal vacuum surrounding generative AI reveals, the same battle is being replayed on a global scale. The US Copyright Office ruled in January 2025 that "prompts alone are insufficient to establish authorship"—a formulation that echoes, almost two centuries later, the idea that pressing a button does not make you a creator. The Andersen v. Stability AI case, which will go to trial in September 2026, tests the very legality of training models on protected works. More than seventy lawsuits now target generative AI companies. The fragmentation between American, European, and Japanese doctrines—maximum permissiveness in Tokyo, strict regulation in Brussels, case-by-case jurisprudence in Washington—adds a degree of complexity that photography, a primarily European phenomenon, did not experience.
The difference, once again, is speed. Photography had fifty years for the law to adapt. Generative AI does not have that luxury. The market is worth $3.2 billion in 2024, with projections of $40.4 billion by 2033 according to Market.US. The European AI Act enters its application phase in August 2026 with fines potentially reaching 3% of global annual turnover. US court decisions expected in the next twelve months will outline the legal framework for an industry already worth tens of billions. Never has the law run so fast behind technology.
The Social Divide, or the Return of the Luddites
Acceleration has a human cost that historical parallels must not obscure. A survey by the Society of Authors UK, published between 2024 and 2025, reveals that 26% of British illustrators have lost contracts directly due to generative AI. A CVL Economics study estimates that 118,500 jobs could be cut in the US entertainment industry by the end of 2026. These figures are not abstract projections. They measure the ongoing destruction of a creative ecosystem.
As documented in our investigation into the cultural war surrounding artistic AI, the structure of this resistance recalls that of the Nottingham Luddites in 1811. Skilled artisans see a technology destroy the value of their expertise, while the owners of this technology invoke progress and the inevitability of history. However, the artists of 2026 possess weapons that weavers did not. The Glaze and Nightshade tools, developed by Professor Ben Zhao at the University of Chicago, have been downloaded eleven million times. They allow for sabotaging the training data of AI models—a form of technical resistance unprecedented in the history of conflicts between art and technology.
Photography caused a comparable trauma among 19th-century portrait painters, whose clientele evaporated in a generation. But this destruction gave rise to one of the most fertile periods in art history. Freed from the constraint of faithful representation of reality—the camera handled it better than they could—painters invented Impressionism, then abstraction. Art reinvented itself because it was forced to do so. It remains to be seen whether this pattern of creative destruction will recur with AI, and how quickly.
The Tool vs. Art: The Eternal False Question
Every technological breakthrough in the history of visual creation has generated the same question. Is the camera art? Is the television screen art? Is the prompt art? The question is poorly posed. It always has been.
The camera has never been art—it is a tool that allows an artist to create. The television has never been art—it is a medium that Paik transformed into aesthetic raw material. The image generator is not art. It is an instrument whose uses range from the mundane to the sublime, depending on who employs it and for what purpose.
The resolution of this debate invariably follows the same trajectory. Technology is first perceived as a threat, then as a tool serving art, finally as a medium in its own right. The "Augmented Intelligence" auction organized by Christie's in February-March 2025 marks the shift to the third phase. The first sale entirely dedicated to AI art at a major auction house, it totaled $728,784 in sales, with 48% of buyers being millennials or Generation Z and 37% new clients.
As our mapping of the AI art market analyzes, artificial intelligence attracts an audience that did not exist in the traditional art world. It doesn't replace a market. It creates a new one, just as photography created an image market unparalleled in the world of painting.
Nicole Sales Giles, a specialist at Christie's, summarizes this perspective: "AI learns everything to create something new. It's influence. Not theft." Karla Ortiz, artist and plaintiff in the Andersen v. Stability AI class action, responds: "Christie's simply doesn't care about artists." The two positions coexist, irreducible to each other. They already coexisted in the 19th century, when defenders of photography and academic painters held equally strong positions.
History suggests that both sides were partly right: photography devastated certain artistic professions, and it enriched the field of creation. The two things do not contradict each other. They are juxtaposed.
The Model Collapse, or the Unexpected Return of Aura
Perhaps the most illuminating historical parallel is not the one we expect. The research by Ilia Shumailov and his co-authors (Nature, 2024) on the phenomenon of model collapse—the degradation of models—introduces a twist that neither photography nor video experienced. The principle: AI models trained on data generated by other models gradually see their performance degrade, leading to collapse. The first manifestations are expected between 2025 and 2026, with a potentially massive impact by 2028-2030.
If this theory is confirmed—and initial data points in this direction—its implications for the art world are profound. Model collapse would create forced scarcity in algorithmic production. Models unable to renew themselves without fresh human data would see the quality of their output plummet. Human art, an irreducible source of original data, would regain a value that the generative tidal wave had eroded. Walter Benjamin, without knowing it, laid the theoretical framework: if aura disappears with infinite reproducibility, it mechanically returns when this reproducibility reaches its limits. Model collapse would be the return of aura through the back door.
This scenario would redefine market hierarchy. Collectors who bet on "pure" AI art—generated without human intervention—would be taking a risky gamble if output quality degrades. Conversely, "AI-augmented" art, where artificial intelligence serves as a tool for a human vision—the practice of Herndon, Chung, Paglen—could see its value grow. The tool would be revalued precisely because it would remain dependent on the hand that guides it. Photography experienced an analogous shift when, after decades of industrial reproducibility, the original print signed by the artist regained a value of scarcity. AI could follow the same path—in thirty years instead of one hundred.
What Museums Tell Us About the Future
The speed of AI art's institutionalization is the most reliable marker of its trajectory. MoMA integrates Anadol into its permanent collection in 2023. The Getty Museum acquires its first AI photograph in January 2025. The Centre Pompidou buys an NFT by Robert Alice in 2024. The "Machine Love" exhibition at the Mori Art Museum in Tokyo attracts a remarkable attendance. The "Creative Machine" exhibition in Beijing, from November 2024 to February 2025, confirms institutional adoption on a global scale.
These acquisitions are not whims of curators seeking modernity. They reflect a precise institutional calculation: museums that ignored photography in the early 20th century took decades to catch up. Those that neglected video art in the 1960s had to fill major gaps in their collections. Institutions that ignore AI art in 2026 risk repeating this error—with a much higher future acquisition cost.
DATALAND, Anadol's museum in Los Angeles, represents an additional threshold. When an artist founds an institution dedicated to their medium, the medium ceases to be a curiosity and becomes an autonomous field. Nadar opened his photographic studio at 35, Boulevard des Capucines in 1860; it was in these same premises that the Impressionists held their first exhibition in 1874. The artist-entrepreneur who builds the infrastructure of their art accelerates history.
Anadol is doing today what Nadar did one hundred and sixty years ago, and what Paik did by transforming television into artistic raw material. The difference: the leap between the first institutionalized AI artwork and the first dedicated museum was accomplished in less than three years.
Neither Apocalypse, Nor Utopia
The most common error in analyzing the relationship between technology and art is to think in binaries. AI will kill art. AI will save art. Photography was not meant to kill painting nor to save it. It destroyed certain practices—miniature portraiture, documentary painting. It revitalized others—Impressionism was born in part from the painter's liberation from faithful representation. And it created an entirely new field.
Video art did not replace sculpture. It coexists with it. Oil painting is doing well in 2026, two centuries after the invention of photography.
Generative artificial intelligence will likely follow the same pattern of coexistence. Human art forms will be weakened—commercial illustration is already under massive pressure. Others will be stimulated: the value of original painting and drawing could grow as generated images become commonplace. This phenomenon precisely reproduces the dynamic by which the scarcity of handmade goods currently feeds an expanding market for art crafts. A new creative field will emerge, with its codes, masters, institutions, criticism—a field of which DATALAND may represent the founding act.
The historical adoption curve suggests that the full integration of AI art into the artistic canon will take thirty to forty years—around 2050. By then, the current debate will seem as strange as that of 19th-century academicians refusing to consider photography as art. But by then too, model collapse may have reshuffled the cards, courts will have settled copyright issues, and the market will have established its own valuation and sorting mechanisms.
The Clock Is Ticking Faster
What the history of photography, video art, and digital art teaches us is not that artificial intelligence will inevitably be accepted as an art form. The process of integrating a new medium follows predictable stages—rejection, exploration, institutionalization, coexistence—and it accelerates with each iteration. Photography took a century. Video, half a century. AI may take a generation.
This acceleration has a consequence that artists, collectors, legal professionals, and institutions must factor into their calculations: decision windows are shrinking. Those who ignore artistic AI today do not have the century of respite that photography's detractors enjoyed in 1859. The legal decisions of 2026—the Andersen v. Stability AI trial in September, the application of the European AI Act in August—will set precedents for decades. Institutional acquisitions in 2025-2026 are already defining the contours of the future canon. The market, with its 37% new clients at Christie's, is being built now.
The philosopher and the weaver, the curator and the coder, the collector and the furloughed illustrator are all experiencing the same historical moment, each from their position in the ecosystem. Baudelaire thought photography was the handmaid of the arts. He was not wrong about the nature of the tool. He was wrong to believe that the tool could not become an art. Artificial intelligence is a paintbrush. The question is not whether it is art. The question—the only one that matters—is which artists, under what conditions, with what rights and what remuneration, will seize it to create what we are not yet able to imagine.
This article concludes the special report "Artificial Intelligence and Creation: The Great Recomposition." Also read: [The Invisible AI Art Market: Anatomy of a Booming Sector], [Portraits of Artists Who Refuse the Binary Debate], [The $300 Billion Legal Void], ["No AI Art": An Investigation into a Cultural War Beyond Art].
