Good Artists Copy, Great Artists Steal

Long before AI art, Steve Jobs said it in 1996, on camera, and credited it to Picasso. Since then, the line has become one of the most repeated phrases in artists and creative culture. It has opened design keynotes, closed a thousand LinkedIn posts, and hung on studio walls as a kind of permission slip. But Picasso almost certainly never said it.

That detail is not a footnote. It is the actual subject of this piece. So before asking what the quote means for AI art, it is worth asking where it comes from. And why a man as protective of his own image as Jobs chose to build part of it on somebody else’s line.

The oldest known version

The trail does not start with Picasso. The earliest documented precursor is nearly a century older, buried in an 1892 issue of “The Gentleman’s Magazine.” In an essay titled “Imitators and Plagiarists,” critic W. H. Davenport Adams drew a line between two behaviors. Imitation, in his view, was legitimate craft. Theft was not. He illustrated the distinction through Alfred Tennyson, showing how the poet had built his verse on the work of his predecessors without ever being accused of plagiarism. The imitation was skilled enough to stay invisible.

The idea sharpens considerably in 1920, when T. S. Eliot wrote what remains the most precise, and least quoted, version of it:

“Immature poets imitate; mature poets steal; bad poets deface what they take, and good poets make it into something better, or at least something different. The good poet welds his theft into a whole of feeling which is unique, utterly different from that from which it is torn.”

Eliot’s version explains something the popular one leaves out: why stealing works at all. The act of taking is not what separates a bad poet from a good one. What happens to the material afterward is what separates them. A poet who defaces it is bad. A poet who welds it into something new is good. In Eliot’s sentence, theft is only half the idea. Transformation is the other half, and it is the half that gets cut every time the quote gets shorter.

Between Eliot and Jobs, the line passes through a haze of secondhand credit. A 1986 text on document preparation attributes “Lesser artists borrow; great artists steal” to Igor Stravinsky, though no primary source for this has ever surfaced. Lionel Trilling and William Faulkner have each been credited with their own versions at different points. By the time the phrase reaches Picasso, it has already been rewritten by several people who may never have said it.

Why Jobs claimed it

Jobs did not invent the misattribution, but he is the reason it stuck. The line first appears publicly in a 1988 “Sydney Morning Herald” article on the Macintosh’s development. It gets its most quoted delivery in the 1996 PBS documentary “Triumph of the Nerds,” where Jobs says on camera: “Picasso had a saying — ‘good artists copy, great artists steal’ — and we have always been shameless about stealing great ideas.”

The context matters here. Jobs is talking about the Macintosh’s graphical interface, developed after Apple engineers visited Xerox PARC and saw a windows-and-mouse system Xerox had no plan to commercialize. He was not being caught out. He was getting ahead of the accusation. Attributing the line to Picasso did two things at once. It borrowed the credibility of a universally recognized genius, and it re-framed “we copied Xerox” as part of a lineage running through art history rather than a shortcut. It is a founding myth dressed up as an art-history lesson.

There is an irony here. The same man who built part of his legend on a misquoted line about stealing was, at the same time, aggressively suing competitors for allegedly stealing Apple’s own interface. So “good artists copy, great artists steal” was never a principle Jobs applied evenly. It was a story he told about his own borrowing. Not one he extended to anyone borrowing from him.

What stealing a language actually means

Strip away the Jobs mythology, and the quote still points to something real about how creative work gets made. No artist invents a visual language from nothing. Every one of them starts by absorbing what already exists: a palette, a compositional habit, a way of framing a face or a landscape. The first attempts are, by definition, imitation.

So what separates an artist who stays derivative from one who does not? Influence alone does not decide it, since influence is unavoidable for everyone. What decides it is whether that influence gets metabolized into something with its own internal logic, or stays visible as a borrowed costume. This is the distinction Eliot made in 1920, before the quote got compressed into a slogan for keynote slides: the theft has to be welded, not pasted on. A style becomes someone’s own not because they hid where it came from, but because they took it somewhere its source never went.

Who actually did this

Jean-Michel Basquiat built an entire visual vocabulary out of things that were not his. Anatomical diagrams lifted from “Gray’s Anatomy.” Copyright symbols pulled from commercial packaging. Saints and phrases borrowed from graffiti culture. None of these elements were his invention. What was his was the collision: the specific, recognizable way he forced unrelated references onto the same canvas until the combination itself became a signature.

Quentin Tarantino has never hidden his sources. He has built a career on saying so out loud. “Kill Bill” openly stitches together Shaw Brothers kung fu choreography, Sergio Leone’s spaghetti westerns, and Japanese samurai cinema. The theft is not disguised. It is the entire selling point. What makes the result his is a directorial voice: pacing, dialogue rhythm, structural non-linearity, reorganizing borrowed genre material into something none of the source films resemble on their own.

Andy Warhol went further than either of them. He did not transform his sources stylistically at all. The Marilyn Monroe silkscreens come almost verbatim from a studio publicity photograph. The Campbell’s soup cans reproduce the original packaging design nearly unchanged. Here, the theft is closer to appropriation in its most literal form. And it became one of the most recognizable signatures in twentieth-century art for exactly that reason: the mechanical repetition was the idea, not a way of disguising one.

DJs and the sampling tradition operate on a version of theft with no ambiguity about its source. A DJ does not write the record they are playing. What they own is the arrangement: which fragment gets looped, what it sits against, how long it plays before the next layer arrives. Entire genres, from hip-hop to house, were built by treating other people’s finished recordings as raw material rather than finished statements.

Across all four cases, the pattern holds. The source material is rarely hidden, and the value never sat in concealment. It sat in what happened to the material after it was taken.

Where the comparison breaks down for AI art

The historical parallel is satisfying, but it needs a harder edge than most versions of this argument give it.

Basquiat referencing “Gray’s Anatomy,” Tarantino referencing Shaw Brothers cinema, a DJ sampling a four-bar break: in each case, what gets taken is a language. A visual grammar, a genre convention, a rhythmic pattern. No single, specific artwork is being reproduced or absorbed wholesale. Warhol comes closest to an exception. But even his source material, a studio publicity photo, a grocery product label, sat in a different legal and cultural category than a living artist’s original, uncommissioned work.

AI image models complicate this in a way none of these precedents fully anticipated. Training a model on a scraped dataset can mean absorbing not the language of thousands of artists, but their specific works. Often without consent, without credit, and without compensation reaching the artist whose paintings taught the system what a brushstroke or a color palette looks like. That is a different act than a painter studying influences in a museum, or a DJ crate-digging for a sample to license.

Theft has always been part of art history, so calling AI art a theft is not the mistake. The mistake is treating it as a single accusation, when it is really two different ones this piece has tried to keep separate. Taking a style, which every artist above did without apology. And absorbing someone’s actual, specific, unlicensed work into a system that can later echo it at scale. Basquiat never needed permission to paint in the vocabulary of artists he admired. Neither did Tarantino, nor any DJ who ever touched a sampler. None of them lifted a finished, individual work whole.

That is the distinction AI art still has to earn, dataset by dataset, model by model. Not whether influence is legitimate. Art history already settled that question. But whether the mechanism of absorption respects the difference between a language anyone can speak and a work that still belongs to someone.

More essays on AI art, authorship, and the questions the field has not settled yet, on partfaliaz.com.