In November 1811, in Nottinghamshire, stocking makers armed with hammers smashed the stocking frames that threatened their livelihood. Two centuries later, thousands of illustrators are flooding ArtStation with visuals emblazoned with the slogan "No To AI Generated Images." The tools have changed. The structure of the conflict has not. Skilled artisans see new technology destroy the value of their expertise; the owners of this technology invoke progress and historical inevitability. Except this time, the Luddites have PhDs, lawyers, and eleven million downloads of digital weapons.
The painful figure
The Society of Authors UK survey (2024-2025) provides a stark diagnosis: 26% of British illustrators surveyed have lost jobs due to generative AI. Among translators, it's 36%. Nearly 78% of illustrators surveyed anticipate a negative impact on their activity.
The United Kingdom is no exception. In a study published in January 2024, CVL Economics estimates that 21.4% of jobs in the US entertainment sector could be displaced by AI by the end of 2026—that's 118,500 positions. Three-quarters of companies in the sector have already used generative AI to eliminate or reduce positions. Research from the Stanford Graduate School of Business (2024-2025) has documented a sharp decline in the market for human-created images after the advent of generators. Their data establishes a direct causal link between the supply of AI and the decrease in demand for human-created art (see our article on the recomposition of the art market in the face of AI).
This is not a hypothetical future. It is a measurable present.
December 2022: The first digital barricade
The revolt took shape on ArtStation, a leading showcase for professional illustrators, concept artists, and designers. In December 2022, thousands of them coordinated an unprecedented move: flooding the site with visuals bearing the message "No To AI Generated Images," transforming the homepage—ordinarily a gallery of graphic prowess—into a uniform wall of protest. This was the first mass demonstration of the creative world against generative AI.
The momentum has not waned. The #NotMyAI petition, launched against the non-consensual use of works to train models, has surpassed 150,000 signatures. In February 2025, when Christie's announced a sale dedicated to AI-generated art, an open letter signed by over 6,000 artists, gallerists, and academics denounced what they perceived as an institutional legitimization of creative plundering. A survey conducted in September 2025 among the artistic community shows that over 90% of respondents view AI-produced art negatively.
A nearly unanimous rejection—rare in a professional environment usually fraught with school rivalries and aesthetic disputes. Generative AI has achieved what few topics have: uniting the creative community in opposition.
Glaze, Nightshade: when artists make their own weapons
Faced with the industrial aspiration of their works by training models, creators did not just protest. They armed themselves. The arsenal emerged from a computer science laboratory at the University of Chicago, under the direction of Professor Ben Zhao.
Glaze works like digital camouflage. Applied to an image before publication, it introduces perturbations invisible to the human eye but which scramble the model's weights during scraping. A generator attempting to learn the style of an artist protected by Glaze will only obtain a degraded, unusable result. Nightshade, its offensive twin, goes further: instead of masking, it poisons. Corrupted data, on a large scale, causes the collapse of the capabilities of the model that ingests it. For Ben Zhao, the challenge is to give artists control over their own work.
The two tools total approximately eleven million downloads—8.5 million for Glaze, 2.5 million for Nightshade. The number is massive. A significant portion of the global creative community has chosen technical action, transforming each published image into a trap for training systems. The Luddite analogy reaches its limit here: the Nottingham stocking makers destroyed machines; the artists of 2025 sabotage the raw material these machines consume. A more subtle, and potentially more effective, strategy.
The betrayal from within
If there is one actor artists target with particular anger, it is Adobe. The Californian company—whose Photoshop, Illustrator, and InDesign have been the daily toolkit for visual creators for three decades—launched its own image generator, Firefly, presenting it as "ethical," trained only on licensed content.
The reality turned out to be murkier. Investigations showed that Firefly had been trained on images from Adobe Stock contributors without their explicit consent. According to community estimates, nearly 62% of the content on Adobe Stock is now AI-generated. For thousands of photographers and illustrators who had supplied the image bank for years—sometimes as their main source of income—the pill is bitter. The tool that replaced them was fed by their own work, by the very company that sold them digital brushes.
A structural paradox. The platforms that hosted creators and profited from their production have become the vectors of their replacement. The content provider has unwittingly become a training data provider. The legal framework, as our analysis of regulatory gaps shows (see our article on the legal uncertainty surrounding AI and copyright), has not yet caught up with this shift.
Platforms facing the wall
The world of creative platforms has fractured. First, the pioneers of refusal. Newgrounds, a historic site for independent animation, banned AI-generated content as early as September 2022, even before the public debate exploded. Getty Images followed by banning AI image uploads—a decision as much commercial as political: the legal risk associated with copyright made these contents toxic for a company built on photographic licensing.
At the other end of the spectrum are platforms that attempted a middle ground. DeviantArt, a historic community for digital artists, added an option in late 2022 allowing users to refuse to have their works used for model training. An opt-out, not an opt-in—by default, works remained available for scraping. The nuance was not lost on anyone.
This mosaic of contradictory policies reflects the absence of an industry standard. Each platform navigates between the pressure from its creators, the demands of its investors—often fascinated by AI's commercial potential—and a regulatory framework that is slow to materialize. An artist's protection depends on the site where they publish.
The union front: one victory, some defeats
The SAG-AFTRA (Screen Actors Guild-American Federation of Television and Radio Artists) strike, which paralyzed Hollywood for part of 2023, set a precedent. Among the central demands was to protect actors against the reproduction of their image and voice by AI systems. The union secured contractual guarantees—a studio cannot create an actor's digital double without their explicit consent and specific compensation.
This victory was due to a particular balance of power. SAG-AFTRA represents 160,000 members in an industry concentrated around a few major studios. Without actors, no films. The leverage is clear, measurable, immediate.
For illustrators, there is nothing similar. The Society of Authors documents job destruction but does not have the same bargaining power. Illustrators work freelance, dispersed, without a single employer to constrain. Their clients—publishers, advertising agencies, video game studios—can discreetly replace a human provider with an image generator. No picket lines form in front of their offices. The atomization of the market makes classical union mobilization almost inoperative.
The SAG-AFTRA model inspires. It does not translate directly.
Art schools in a fog
The American National Art Education Association raised the alarm in April 2024: art schools are "scrambling" to adapt their curricula. The observation is shared in Europe. What should be taught to a first-year illustration student when photorealistic image generators are accessible to everyone for a few euros a month?
The question involves concrete pedagogical choices. Some schools integrate AI as a tool, just as Photoshop was twenty-five years ago: prompt engineering, assisted artistic direction, post-processing of generated images. Others strengthen the demand for manual techniques and traditional drawing, betting that the value of human know-how will increase as it becomes rarer.
Neither of these approaches is based on certainties. Schools are training students for a market whose configuration in five years no one can predict. Relying exclusively on AI integration risks training tool operators who will themselves be automated in the next technological iteration. Refusing all adaptation risks producing excellent but unemployable artisans.
The misunderstanding of collectors
The collecting world, often perceived as distant from these battles, sends a clear signal. The 2024 Hiscox survey reveals that 82% of art collectors want mandatory labeling of AI-produced or AI-assisted works. They want to know what they are buying. They want the distinction between human creation and algorithmic generation to be visible, traceable, guaranteed.
The discrepancy with a part of the institutional market is striking. Christie's organizes a sale dedicated to AI art; 82% of collectors demand transparency. Neither auction houses, nor galleries, nor online platforms have adopted a common labeling standard. A grey area thrives, where generated images circulate without mention of their origin.
The real conflict: who owns creative work?
Beneath the surface of petitions, protection tools, and market restructuring, an older question is being replayed. It concerns not technology as such, but the ownership of work. Artist Karla Ortiz, a plaintiff in the class action lawsuit Andersen v. Stability AI—one of the most closely watched trials in the sector—embodies this front line. Her argument can be summarized in one sentence: a model trained on millions of works without authorization produces derivatives of these works without compensating their authors. The technological benefit relies on uncompensated extraction of creative value.
The argument is not just legal. It is economic. Stanford's research has shown that the entry of generative AI into the market mechanically causes a drop in demand for human images. Not an indirect or diffuse effect: a measurable, documented replacement. The 26% of British illustrators who lost contracts are not victims of a slow evolution of taste. They are victims of direct substitution by a tool powered, in part, by their own previous work.
This circularity—being replaced by a machine trained on one's own work—fuels a rage that statistics alone cannot capture. It explains why the anti-AI movement in art is not like ordinary resistance to change. In the creators' perception, something is akin to theft—even before the courts have ruled.
A forcibly reorganized ecosystem
Platforms are divided between those that ban and those that integrate. Unions with blocking power—SAG-AFTRA—secure protections; others witness the erosion of their ranks. Art schools rewrite curricula without knowing what world they are training for. Collectors demand transparency that no one can guarantee on a large scale. Glaze and Nightshade prove that a technically competent community can retaliate, but their effectiveness depends on massive adoption against training corpora numbering billions of images.
Comparing this moment to the Luddite revolts is not just a rhetorical exercise. The stocking makers of 1811 lost. Textile industrialization swept away craftsmanship in a few decades, and mechanical looms triumphed. But the Luddites, through their resistance, also forced the adoption of laws on child labor, factory conditions, and machine regulation. Their economic defeat produced social advances.
The "No AI Art" movement could follow a similar trajectory. It may not prevent generative AI from transforming the illustration market. But it has already achieved concrete results: union protections in Hollywood, technical defense tools, growing regulatory awareness—slowly—in Europe and the United States, bans on certain platforms.
The 118,500 jobs that CVL Economics projects will disappear by the end of 2026 are not an abstraction. They are illustrators who stop receiving commissions, concept artists whose positions are not renewed, translators replaced by automated pipelines. The next chapter will be written in courts, in parliaments, and on the screens of millions of creators who, every morning, open their drawing software wondering how much longer this gesture will have market value.
This article is part of the dossier "Artificial Intelligence and Creation: The Great Recomposition". Read also: [The Art Market Facing AI: Mapping a Disruption] and [Copyright and Generative AI: In the Legal Void].
