FRAMES
The Myth of the Hand
Authorship lived through the factory; it will live through generative tools
▫️ The hand was never the ground of authorship, only its most legible proxy. From workshops to fabrication studios, art has long been made through delegated and structured execution. Generative tools do not erase the author; they expose the instability of the old signs by which markets and institutions thought they could recognize one.
Art has been produced inside systems: workshops, studios, print rooms, fabrication floors — for longer than the myth of the solitary hand has existed to obscure them. History is rougher than the proxy permits, and more useful for it. In each of these settings, the named artist directed while others executed. Authorship lived not in touch but in governance: in intent, constraint, selection, approval, and the liability carried under one name. The hand performed cultural labour, certainly, but mainly as proof of presence for buyers who preferred a simpler story. What the hand proved was not authorship. It proved proximity, and the two were treated as equivalent for as long as the market had no reason to separate them.
The Renaissance workshop makes the transfer visible; the production records it directly. A master like Perugino or Verrocchio ran an operation with a dozen or more assistants at various levels of competence. The master set the composition, often through a full-scale preparatory drawing — the cartoon — and assigned sections of the panel or fresco according to skill. Faces and hands, the passages where individual judgement was most exposed, were typically reserved for the master or the most advanced pupils; drapery, landscape, and architectural backgrounds were delegated downward. Correction followed: repainting a passage, adjusting a contour, unifying the tonal register across sections painted by different hands. What made the finished work "his" was not that he had touched every surface but that his decisions governed the composition, the colour logic, the figural types, and the standard to which execution was held. When a patron paid for a Verrocchio, the contract specified materials, dimensions, and sometimes the degree of the master's personal involvement, but the name on the contract warranted a production standard, not a forensic record of contact.
Authorship resided in the organizing intelligence. The market understood that well before anyone needed to theorize it. Modern fabrication keeps the same split and states it more plainly. Fabricators execute to specification, print studios pull editions, technicians realize instruction-based works. Warhol scaled production as throughput. Koons separates conception from fabrication under a single signature regime. Hirst's spot paintings move through repeatable studio protocols. Each case demonstrates the principle the workshop established: the named author governs the field of decisions, and the market accommodates that structure so long as accountability can be located — in invoices, fabricator statements, edition records, and studio correspondence. The accommodation is not generosity. It is structural dependence: the market needs the authorship claim to hold because the asset depends on it.
Reading Time
12 MINUTES
Words
GIULIANO CESAR
Header
G—G LAB
Published
MAR 2026
Reading Time
12 MINUTES
Words
GIULIANO CESAR
Header
G—G LAB
Published
MAR 2026
Authorship lived less in touch than in governance:
in intent, constraint, selection, approval, and the liability
carried under one name.
Andy Warhol by G—G Lab.
▲
Authorship lived less in touch than in governance:
in intent, constraint, selection, approval, and the liability
carried under one name.
All images are included for educational purposes only, with copyright retained by their respective creators. Image-related concerns are addressed promptly.
The administrative infrastructure that underwrites these claims will be tested hardest when the hand disappears as a proxy.
Inside the museum's registrar stack, cataloguing standards distinguish primary creators from secondary roles — printers, technicians, workshops — because institutions need attribution that can survive scrutiny. Registrars maintain provenance files tracking title history, transfers, restrictions, and the paper trail that allows an object to hang without immediate legal jeopardy. At the point of sale, auction houses translate those realities into calibrated language: "attributed to" indicates confidence short of proof; "studio of" or "workshop of" marks execution within an authorized system rather than by the artist's hand. These are not hedges. They are pricing instruments. Whether a work enters the room as saleable, lands as discounted, or stalls as dead on arrival depends on which phrase the catalogue carries.
Those systems did not create authorship. They recorded, secured, and defended claims that arose first from artistic practice. But their reliability depended, in most cases, on a visible chain of physical evidence: the hand, the studio, the fabricator's statement, the crate, the invoice. When the hand was the proxy, the infrastructure worked because it could point to something tangible. The question is narrower than the general anxiety about AI suggests: what happens to that infrastructure — not to authorship itself — when the proxy dissolves, when the physical chain that made authorship legible to institutions no longer accompanies the object?
Sol LeWitt stated the model of non-manual authorship without apology. The certificate and diagram identify the work; draftspeople execute under instruction. Authorship sits in authorized rules and permitted procedure, not in hand-applied pigment. Without a certificate, a wall drawing is a wall plus a story; with one, it becomes an artwork that can travel, be insured, and be sold. The work's identity is a designed field of possibilities realized according to rule, and that field is the authored object, not the wall it happens to occupy. The contrast with generative practice clarifies what is missing: in LeWitt, the rule-set is part of the work's legibility; in generative practice, the governing structure often remains private, disposable, or treated merely as production scaffolding unless someone chooses to preserve and disclose it. LeWitt's system works because the administrative infrastructure can read it. Generative practice has not yet given the infrastructure equivalent terms.
The work's identity is a designed field of possibilities
realized according to rule, and that field is the authored object,
not the wall it happens to occupy.
Sol LeWitt by G—G Lab.
▲
The work's identity is a designed field of possibilities
realized according to rule, and that field is the authored object,
not the wall it happens to occupy.
All images are included for educational purposes only, with copyright retained by their respective creators. Image-related concerns are addressed promptly.
The Warhol authentication board exposed the fault line — not in authorship, but in the administrative system that claimed to adjudicate it.
An authentication board is a gatekeeper capable of turning a work into either an asset or a problem, and Warhol's studio model made that gate unusually contentious. The institution was asked to authenticate an artist whose production had diluted the hand by design: assistants, outside printers, mechanical repeatability, authorization as method. The market tolerated that logic for years because it depended on it. When exposure rose, it tightened its stance, and the tightening revealed that the tolerance had always been conditional on low risk, not on settled principle.
The Red Self-Portrait dispute turned that conditionality into a test. A work that had once received early validation was later refused under stricter criteria, favouring demonstrable proximity and evidence of Warhol's direct involvement over earlier, looser forms of approval. The decision was not merely aesthetic. It reset the warranty terms around the object.
What followed was a cost event disguised as a connoisseurship dispute. Simon-Whelan's lawsuit sought more than $20 million under federal antitrust claims tied to the board's denials, arguing that a private authentication gate could act like a market regulator — because a single decision could convert a work from asset to liability. Win or lose, the case repriced the act of saying yes or no. In 2011, as exposure escalated, the Foundation announced that it would dissolve the authentication board at the beginning of 2012; contemporaneous reporting placed legal fees in the millions, with defence costs described as near $7 million. Later litigation continued to contest additional reimbursement beyond policy amounts already paid.
The case is not history. It is precedent. The board could not hold the line between "a Warhol" and "not a Warhol" because the artist's own method had made that line unstable. Once authentication becomes a cost centre, counsel arrives early, standards harden, and authorship is recast as a liability instrument. The infrastructure built to record authorship begins to produce it, or refuse it, under market pressure. That dynamic does not depend on any particular medium. It depends on how much financial exposure a single attribution decision can create. Generative tools, by eliminating the hand entirely, do not introduce this instability; they guarantee that no future authentication regime can fall back on the proxy that Warhol had merely weakened.
Authorship has survived workshops, fabrication,
instruction, repetition, and industrial scale because
it never depended on the hand.
Damien Hirst by G—G Lab.
▲
Authorship has survived workshops, fabrication,
instruction, repetition, and industrial scale because
it never depended on the hand.
All images are included for educational purposes only, with copyright retained by their respective creators. Image-related concerns are addressed promptly.
Generative tools leave authorship intact but break the proxy the infrastructure relied on. A work can arrive visually complete while the witnessing layer — who directed it, what was chosen or refused, what was edited, what was permitted — fails to travel with the file.
In 2023, Boris Eldagsen won the Sony World Photography Awards' Creative category with a DALL·E-generated image and then declined the award, arguing that the competition could not reliably distinguish AI output from photography. The reverse test arrived in June 2024: Miles Astray entered a real photograph, *Flamingone*, in the AI category of the 1839 Awards; it won Third Place and the People's Vote, then was disqualified once disclosed as non-AI. Both cases tested the same administrative infrastructure this essay describes, and both times it failed — because it was still reading the surface for clues the surface no longer reliably carries. Eldagsen demonstrated that a generative image could pass as photographic; Astray demonstrated that a photograph could pass as generative. The symmetry is the point. When the surface cannot certify what produced it, the surface no longer reads as evidence.
Generative workflows sit inside the same authorial structure as every other delegated regime. The chain persists: parameters, prompting rounds, selection, compositing, retouching, acceptance. Human decisions occur at every stage. What changes is not the existence of authorship but the ease with which that authorship can be read from the finished surface. In older studio settings, uncertainty was softened by default witnesses — payroll, assistants, fabrication contracts, shop logs, invoices, correspondence. In private computation, equivalent iteration compresses into software and may leave little inspectable residue. The authorship is no less real. It is less legible to the systems that need to certify it.
The practical difficulty is immediate, not speculative. A curator may have a finished image and a name to attach to it, yet lack a clear record of which tools were used, which permissions apply, or who made the final selections. For physical work, gatekeepers can name the chain because third parties sign statements and studios can be visited. For computational work, institutions often rely on artist assertions about degree of control, source boundaries, and process steps, then decide how much weight those assertions can bear. Attribution language built for the physical studio — "attributed to," "studio of," "workshop of" — has no settled equivalent for the generative pipeline.
That gap is where the infrastructure must develop. Legal thresholds diverge across jurisdictions on copyright, moral rights, consumer protection, and disclosure duties; markets, institutions, and audiences will respond unevenly, and no final settlement is in view. But the problem is administrative, not ontological. Authorship has survived workshops, fabrication, instruction, repetition, and industrial scale because it never depended on the hand. It depended on the organizing intelligence: the capacity to shape a field of choices into coherence, to hold that coherence to a standard, and to accept responsibility for what enters the world under one name. That structure persists. What has collapsed is the proxy that made it easy to read, and with it, the assumption that a surface could serve as evidence of an author. The administrative systems that record, certify, and defend authorship claims are where resolution will be built, if it is built at all. Those systems have adapted before. Whether they adapt quickly enough, and whether the adaptation serves artists or merely protects intermediaries, remains the open question. ◾️
Loved this read?
Here’s more you might
connect with:
When the Camera Rewrote the Painting
And why generative systems are rewriting the camera
FRAMES
What Images Are Not
How visual truth was manufactured long before generative tools
FRAMES
Big Figures Are Writing the Rules
How Christie’s, Sotheby’s, and Phillips are defining generative art
FRAMES
