The Ghost in the Canvas: When AI Art Sells for Millions, Who Really Gets Paid?
by Dr. Zara Osei, AI Culture Lab
In 2018, Christie’s auctioned off an artwork titled Portrait of Edmond de Belamy—a blurry, ghostly image of a fictional man—created by an algorithm. It sold for $432,500, exponentially more than its estimated value. But the real shock wasn’t the price; it was the signature in the bottom-right corner: a snippet of code. The sale triggered a cultural and economic riddle: if an AI creates the artwork, who deserves the check?
Since then, AI-generated art has surged from a novelty to a legitimate force in the art world. Generative adversarial networks (GANs), diffusion models like DALL·E, and multimodal systems such as Midjourney and Stable Diffusion have become the brushes and pigments of a new digital avant-garde. Galleries now host exhibitions of AI-generated works; collectors bid on algorithmic paintings; and tech companies brand themselves as the new patrons of the arts. But amid this frenzy, a haunting question lingers in the canvas’s coded shadows: when AI art sells for millions, who really gets paid?
Whose Work Is It Anyway?
To answer that question, we must first unravel what “authorship” even means in this new landscape. AI-generated artworks are never created by AI alone—they are the product of layered collaborations. There’s the coder who designed the model, the dataset curator who trained it, the prompter who guides the AI to output a particular image, the platform that hosts it, and the gallery that packages and sells it.
In traditional art, the chain of value is clearer: the artist creates, the dealer sells, the collector buys. But in AI art, value is dispersed across a network of human and computational agents. Take Edmond de Belamy: although it was “created” by a GAN, the Paris-based collective Obvious was behind the prompt and presentation. The algorithm itself was based on a pre-existing codebase developed by 19-year-old artist Robbie Barrat, whose work was used without his direct involvement—and without compensation.
This is not just an intellectual property tangle. It is a reframing of creative labor itself. In the AI art ecosystem, authorship is modular. Labor is fractured. And yet, profits often consolidate in the hands of those who can navigate the market—not necessarily those who created the foundational tools or datasets. This raises a crucial issue: are we witnessing the emergence of a new creative class, or the exploitation of invisible labor beneath the aesthetic surface?
The Invisible Workers Behind the Art
Consider the developers behind generative tools like DALL·E, Midjourney, or Runway. While the companies may profit through subscriptions or licensing, the actual coders and dataset curators are often salaried employees, far removed from the auction house spotlight. At the same time, countless images scraped from the web—produced by unknown photographers, illustrators, and digital artists—form the training corpus. Their work is embedded in the AI’s output, yet they receive no royalties, recognition, or recourse.
Then there are the prompt engineers—the new class of creators whose artistry lies not in brushstrokes but in linguistic precision. Some prompts are simple; others are deeply iterative, requiring hours of refinement and aesthetic judgment. Should these individuals be credited as artists? Are they co-authors, curators, or something entirely new? At present, the answer seems to depend less on theory and more on market power.
Galleries and auction houses are now grappling with how to classify and price such works. Some attribute authorship solely to the person who submitted the prompt. Others recognize a collective entity or assign credit to a studio. Rarely, however, is the ecosystem of creation acknowledged in full. The AI itself, of course, cannot own anything—legally or conceptually. But the humans behind it? Their recognition remains uneven at best.
A Market Hungry for Ghosts
The art market has long fetishized the myth of the solitary genius. But AI-generated art undermines this fantasy by exposing the layered, networked reality of modern creativity. And yet paradoxically, the market continues to seek human drama—even in code. Artists like Refik Anadol, who use machine learning in large-scale data-driven installations, craft narratives around their own authorship to frame the work. The market rewards this. Others, like the German collective Studio Drift, mask the computational complexity behind minimalist aesthetics, letting the mystique drive value.
Auction houses and galleries are quick to capitalize on this tension. AI art becomes a spectacle—a ghost in the machine, a product of intelligence both human and artificial. But who actually owns that ghost? In many cases, it’s not the artist or the coder or the prompter—it’s the company with the infrastructure, the brand, the legal muscle. The monetization of AI art reflects a broader shift in the cultural economy: value is migrating from creation to orchestration.
Toward an Ethics of Recognition
We are at a crossroads. AI-generated art is not just a curiosity—it is a mirror, reflecting how we value labor, creativity, and ownership in an algorithmic age. The current system tends to reward those already equipped with market access, platform control, or institutional legitimacy. Those who provide the raw cultural material—often without consent or compensation—remain unseen.
To move forward, we need new frameworks for attribution and compensation. These might include transparent credit systems that map the full chain of creative input, platform-based royalties for prompt engineers and dataset contributors, or cooperative ownership models for AI art studios. Some technologists are already exploring blockchain-based provenance tracking to address these gaps. But without cultural and legal shifts, such tools remain symbolic rather than systemic.
Ultimately, the question is not whether AI can create art. It can. The question is whether our systems of valuation and recognition can evolve fast enough to meet this new reality. If we fail to answer that, the ghost in the canvas will not be the AI—it will be the multitude of human creators whose labor, though crucial, goes unpaid and unseen.
The canvas is no longer blank. It is layered, entangled, and encoded. But we must ask: in this brave new gallery of algorithmic images, whose name is truly written in the frame—and whose signature fades into code?
