By Alexander Stone
There’s a heist happening. It isn’t loud. There are no balaclavas, no sawn-off shotguns, no splintering doors. The silence is the point. It is a heist conducted in the humming vacuum of a server farm, an act of intellectual plunder so vast, so audacious, it almost defies comprehension. Big Tech, in its relentless pursuit of a synthetic mind, has become the world’s most sophisticated cat burglar, and its target is the entirety of human knowledge. The emergent sentience we are promised in these new AI models is being birthed from a stolen library, its consciousness a mosaic of pirated thought. And buried deep within the code, amongst the scientific papers and potboiler novels, is the very soul of the art world – its histories, its manifestos, its visual memory – being quietly stripped for parts. Creativitys.UK will tell you about the ghost in the machine, and how it learned to paint by raiding the museum after dark.
A Shadow Library’s Purpose
At the heart of this story is a notorious, almost mythical entity: Library Genesis, or LibGen. For years, it has existed as a kind of digital samizdat for the twenty-first century, a sprawling, illicit archive of millions of books, academic journals, and texts. For students in developing nations, researchers locked out by exorbitant paywalls, and the intellectually curious on a budget, LibGen has been a lifeline – a grey-market beacon of accessible knowledge. It is a pirate utopia, operating on the principle that information wants to be free, regardless of who wrote it, who published it, or who holds the copyright. Its very existence is a defiant gesture against the commodification of knowledge. But that defiance, that vast and lawless repository, has now become the unwitting fuel for a machine of unimaginable scale and ambition.
The Meta Revelation
The curtain was pulled back not by a whistleblower, but by the dry prose of court documents. Legal proceedings against Meta, the parent company of Facebook and Instagram, unearthed internal communications that were breathtaking in their corporate nonchalance. Senior staff, with the explicit approval of CEO Mark Zuckerberg, decided to feed their flagship AI model, Llama, with the entirety of Library Genesis. They didn’t just dip a toe in; they backed up the digital truck and siphoned the whole lot. Let the sheer scale of that sink in. A multi-trillion-dollar corporation, rather than ethically sourcing or licensing data, simply took it. This wasn’t a scrappy start-up cutting corners. This was a calculated, top-down decision to build the future of artificial intelligence upon a foundation of mass copyright infringement. The products of this plunder are now embedded in the apps we scroll through daily, their algorithms whispering with the stolen voices of countless authors.
Art’s Digital Desecration
Whilst the theft of novels and scientific papers is egregious enough, the ingestion of the art world’s specific canon feels like a different kind of violation. A scan of LibGen’s holdings, cross-referenced with what Meta consumed, reads like a curated tour of a world-class museum’s library, except the librarian is a kleptomaniac. We find Gagosian Gallery’s lavish monograph on the painter Jenny Saville, a book that represents the culmination of an artist’s career and a gallery’s immense reputational and financial investment. We find exhibition catalogues from the Solomon R. Guggenheim Museum for landmark shows on Mark Rothko and the Russian Avant-Garde – these are not just books, they are vital records of scholarship, curatorial vision, and historical context. The list is a dizzying, tragic roll call of cultural theft: Kehinde Wiley’s Brooklyn Museum show catalogue, books by Andy Warhol in four languages, the intimate memoirs of Peggy Guggenheim and Marina Abramović, multiple issues of esteemed journals like The Art Bulletin and The Burlington Magazine. Even the fiercely original voices of the Lebanese-American painter Etel Adnan and the Chinese dissident artist Ai Weiwei have been hoovered up, their words and works assimilated into the machine.
The Machine’s Soulless Palette
This is not simply about text. Art books are visual objects. They contain high-resolution reproductions of artworks, carefully photographed, lit, and colour-corrected. They are arguments made in images as much as in words. When an AI model consumes these, it isn’t just learning about art history; it’s learning a visual language it has no right to. It is learning the brushstrokes of Wassily Kandinsky from his own treatise, Point and Line to Plane. It is learning the principles of humane architecture from pirated translations of Frank Lloyd Wright. It is learning composition, colour theory, and form from the life’s work of generations of artists, all without credit, context, or consent. This is visual larceny on an industrial scale. The AI learns to mimic, to pastiche, to generate images ‘in the style of’ an artist whose soul and struggle it cannot comprehend. It flattens culture into a series of prompts, creating a predictable slurry of reheated ideas. The result is a machine that can generate an infinite number of technically proficient, spiritually empty images, forever trapped in an uncanny valley of its own stolen making.
A Crisis of Authorship
This is not a uniquely Meta problem. Court filings reveal that OpenAI, the creator of ChatGPT, has also supped at the poisoned well of LibGen. This is an industry-wide pathology. The race to build the most powerful AI has fostered a gold-rush mentality where ethics are an afterthought. The very people who create the culture these systems mimic are the ones being erased from the equation. Consider the exquisite irony of Jerry Saltz, one of the world’s most prominent art critics, having his books How to Be An Artist and Art Is Life pirated to teach a machine that may one day render his profession obsolete. Or the discovery that the work of the very journalist reporting on this story was also in the database, consumed by the same entity she was investigating. This is a snake eating its own tail, a system that devours its own creators and critics with chilling impartiality. It’s a profound crisis of authorship. If the sum of all human expression can be digitally annexed and used as raw material, what does it mean to create something original anymore?
The architects of this new world will speak of progress, of innovation, of democratising creativity. But their machine is haunted. It is haunted by the ghosts of every book it consumed, every artwork it scanned, every line of poetry it parsed. It has a perfect memory and no conscience. It can replicate the style of John Waters’s irreverent wit or David Hockney’s Californian cool, but it will never understand the human impulse that drove them to create. The Llama model may learn to write and to draw, but it is built on a lie, a grand and silent theft that targeted the very heart of what makes us human: our stories, our images, our art. And we are left to wonder what kind of future we are building when its foundation is laid with stolen ghosts.
