Few stories in the art world are more irresistible than the miraculous rediscovery. A masterpiece turns up in an attic or storage locker, bought for next to nothing, overlooked for years and now potentially worth millions. The ingredients are perfect: luck, secrecy, scholarship, money and the possibility that everyone else missed something extraordinary.
The Basquiat exhibition that opened at the Orlando Museum of Art in 2022 had all these ingredients. Twenty-five works on cardboard were presented as long-lost creations by Jean-Michel Basquiat, supposedly made in 1982, sold directly to television screenwriter Thad Mumford and stored away for decades. Their appearance promised to add a substantial group of unknown works to the brief career of one of the most famous artists of the late twentieth century. Instead, the exhibition became the centre of an art scandal. The FBI seized the works, and a former auctioneer later admitted that he and an accomplice had created the fakes.
Yet a question lingered around the edges of the story. Even if some of the group were obvious fakes, could genuine Basquiats have been mixed in among them?
That was the puzzle considered in this episode of Is It? My guests were Carina Popovici, founder and CEO of Art Recognition, and Magnus Resch, one of the world’s best-known art-market commentators. Resch is a bestselling author, entrepreneur and economist, with a gift for exposing the behaviours, incentives and irrationalities that drive the art market.
He has also launched an eponymous app, Magnus, intended as a kind of Shazam for art. Photograph an artwork and the app identifies the artist and title, supplies price information and comparable sales, and recommends galleries and museums according to the user’s interests. Magnus and Art Recognition approach the market from different ends, but share a premise: greater transparency creates greater confidence.
The Basquiat case showed just how badly that transparency is needed.
1. The art market wants to believe in miracles
The first great insight from our discussion was that the appeal of the Orlando works cannot be separated from the emotional architecture of the story surrounding them.
As Magnus put it, everyone wants the overlooked treasure. Collectors may be sophisticated and surrounded by advisers, but they are not immune to the fantasy of discovering what everyone else missed. The higher the potential reward, the more seductive that fantasy becomes. A newly discovered Basquiat could be worth millions almost overnight.
Basquiat is particularly vulnerable to this phenomenon. His visual language is instantly recognizable: crowns, skull-like heads, scrawled words, anatomical fragments, emphatic lines and deliberate crossings-out. He is globally famous, his market is enormous and his career was brief. That combination creates both scarcity and demand. It also encourages people to believe that unknown works must still be waiting somewhere.
High prices therefore produce a paradox. They encourage scrutiny, but also dubious discoveries, because the possible reward is so great. Every storage locker or forgotten family collection begins to look like a lottery ticket. Magnus noted that the stronger an artist’s market becomes, the more questionable works emerge bearing that artist’s name.
The problem is intensified by gaps in the authentication system. Many artist estates stopped offering opinions after costly legal challenges, while catalogue raisonnés may be incomplete or contested. In that vacuum, a persuasive story can acquire more weight than it deserves.
The Orlando collection arrived in exactly this space between desire and verification. It offered a tale people wanted to believe: the young Basquiat, not yet world-famous, rapidly producing works and selling them cheaply to someone who happened to preserve them. The narrative did not merely accompany the paintings. It helped make them plausible.
This is a recurring danger in authentication. People think they are judging the object, but they are often judging the story—and a good story can lower defences.
2. Museums do not authenticate works, but they create legitimacy
The second major takeaway concerned the role of institutions. Museums are not, strictly speaking, authentication bodies. Yet inclusion in a museum exhibition is widely treated as a form of approval.
As Magnus observed, museums do more than display art. They create legitimacy. Once a work hangs in a respected institution, it appears to have passed through curatorial and scholarly scrutiny. The museum’s name enters its exhibition history and can later be cited by dealers, auction houses and owners. Uncertainty declines, confidence rises and value can rise with it.
That is why the Orlando exhibition mattered so much. The museum did not simply provide a room. It gave the works a place in the historical record. Visitors encountered them as Basquiats; photographs and articles circulated. Had the exhibition proceeded, the museum’s imprimatur could have followed them for years.
This is a kind of laundering of uncertainty. A questionable work begins in an online marketplace, changes hands, accumulates paperwork and appears in a publication, gallery or exhibition. Each new context makes it look less like an unsupported claim and more like an established object.
Carina noted that some of the disputed Orlando works were reported to have circulated through eBay. It is tempting to dismiss such platforms as a separate, low-end world whose fakes can only deceive the naïve. But dubious works do not necessarily remain where they first appeared. They move. What begins as an eBay problem can become a gallery problem, an auction-house problem and, eventually, a museum problem.
By the time a doubtful object reaches a museum, owners, intermediaries, curators and institutions all have something to lose. An investigation that might once have been routine becomes adversarial.
The Orlando scandal therefore illustrates a broader principle: authentication is most useful before reputational and financial commitments harden. The ideal moment to test a collection is not after the press preview, the opening gala or the FBI raid. It is before the exhibition contract is signed.
3. Art Recognition found no hidden Basquiat among the group
For the episode, Art Recognition analyzed digital images of twenty-seven painted surfaces associated with the twenty-five physical works; two of the cardboard pieces were painted on both sides. Every result was negative. The model assessed the images as non-authentic with probabilities ranging from 82 to 99 percent.
There was, in other words, no AI evidence supporting the theory that genuine Basquiats had been mixed among the obvious fakes. The results were not narrowly divided or equivocal. They were consistently and strongly negative across the group.
Basquiat’s style might seem difficult for an AI model: spontaneous, unstable and unfinished. An imitator can reproduce crowns, heads, fractured words and agitated lines. But the model is not simply searching for Basquiat-like symbols. It analyzes how an image is constructed.
Carina described Basquiat’s apparent wildness as a potential advantage. His fast decisions, hesitations, revisions and crossings-out contain traces of personal execution. A forger may imitate the visible vocabulary, but it is far harder to reproduce the entire chain of choices that created the work. The distinction is between copying the appearance of spontaneity and actually working with the artist’s characteristic rhythm.
That makes AI image analysis useful as an early warning system. It does not replace provenance research, connoisseurship or material analysis, but it can screen a large group quickly and indicate whether deeper investigation is justified.
In the Orlando case, twenty-five works constituted a major proposed addition to Basquiat’s oeuvre. Testing them before exhibition would have required only digital images and could have raised immediate, serious concerns. Instead of serving as a late tiebreaker in an already explosive dispute, AI could have been one of the first lines of defence.
Magnus imagined the logical future of this process: before a work is purchased, insured, exhibited or offered at an art fair, it is uploaded for preliminary verification. The result does not end the investigation, but helps determine whether the object deserves to proceed to the next stage.
The episode ultimately returned to a simple piece of advice. When the upside is enormous but the evidence is mixed, responsible collectors and institutions should focus first on the downside. What happens if the work is not genuine? What would that mean financially, legally and reputationally?
The best collectors, Magnus suggested, are often defined as much by the deals they refuse as by those they complete. The same might be said of museums. In a market built partly on hope, restraint is an undervalued form of expertise.
And when a discovery seems almost too good to be true, the right response is not to stop being curious. It is to become more curious, earlier.
You can listen to the whole episode on Spotify or Apple podcasts.