From forgery to artificial intelligence, Professor Noah Charney reflects on the evolving methods of authentication, the persistence of art crime, and the future of attribution in a rapidly changing technological landscape.
Professor Noah Charney is an internationally recognised art historian, author, and leading authority on art crime and forgery. He is the founder of ARCA — the Association for Research into Crimes against Art — the pioneering interdisciplinary organisation dedicated to the study and prevention of art crime, illicit antiquities trafficking, and cultural theft.
Over the past two decades, Charney has helped establish art crime as a serious academic and investigative field, frequently advising museums, collectors, law enforcement bodies, and media organisations on issues surrounding attribution, theft, forgery, and provenance. His work has been featured in publications including The New York Times Magazine, and he is widely regarded as one of the foremost public intellectuals working at the intersection of art history, criminality, and emerging technology.
He is the author of numerous books, including The Art of Forgery, considered one of the definitive contemporary studies of forged art and the mechanisms through which fakes enter the market. His research explores not only the psychology and methods of forgers, but also the institutional vulnerabilities that allow fraudulent works to circulate at the highest levels of the art world.
Alongside his academic work, Charney has emerged as a prominent voice in debates surrounding artificial intelligence and authentication. In advisory roles with firms such as Art Recognition, he has examined how AI image analysis is reshaping attribution practices and challenging long-standing assumptions about expertise, connoisseurship, and authenticity.
His recent projects include Is It? The Art Mystery Podcast, exploring attribution mysteries through AI-assisted analysis, bringing together scholars and specialists to reassess contested artworks and uncover new discoveries within art history.
In conversation with The Exclusivist, Professor Charney discusses the future of AI authentication, the persistence of forgery in elite markets, the dangers of the “Provenance Trap,” and why the art world has yet to fully embrace the technologies capable of transforming attribution and fraud detection.
Interview with Professor Noah Charney
On his foundational work
You founded ARCA, the Association for Research into Crimes against Art. What are the most significant shifts you’ve seen in art crime and forgery over the past decade, particularly at the upper end of the market?
I founded ARCA way back in 2006. It was the first nonprofit research group that brought together the relatively few people around the world who focus their study on art crime. I was very lucky because, in the same period of time, I was featured in The New York Times Magazine as having effectively established the study of art crime as a new field of study, and that really kick-started things for me. I’ve been the go-to art crime professor since then.
Over the last 10 years, one of the most significant shifts we’ve seen is the increased ability of technology to detect forgeries and flag looted or stolen works through image analysis. If an object comes onto the market, we now have software technology capable of conducting image-based searches to see whether an image advertised online—on platforms like eBay, for example—matches any known work registered as stolen. This works very well if a work was stolen from an existing collection and reported to police.
The problem is that it doesn’t really work with looted antiquities, because those objects were taken directly out of the earth, or sometimes the sea, before anyone had a chance to catalogue them. They never appear on any stolen art registry or database, so the software doesn’t know what to look for. We also now have incredible abilities through material forensic testing and AI image analysis to determine whether works are authentic and correctly attributed, but it still takes an extra step for somebody to say, ‘Yes, I want to run these tests,’ and the art trade hasn’t fully caught up yet.
There still isn’t enough demand within the market to insist that every object worth at least six figures should undergo standard material forensics and AI image analysis to reassure the buyer that what they think they’re buying is, in fact, what they’re getting. I think that may change in the future, especially because AI image analysis is relatively inexpensive and doesn’t require moving the object. You can test a high-resolution digital image of a painting, and it’s quite definitive. The technology is the main difference. We haven’t seen much change in what’s being stolen or forged, but the capabilities now exist—we just aren’t using them as universally as we should.
On AI and modern attribution
From your perspective working in art crime and authentication—and in advisory roles such as with Art Recognition—how is AI changing authentication at the intersection of improved detection and increasingly sophisticated forgery?
Traditionally, there have been three ways that art was authenticated. I call them the three pillars.
The first is connoisseurship—the expert’s eye—which remains the default way we authenticate works of art and determine attribution based on informed opinion. But experts often disagree, and they can get things wrong. The second is provenance research: the documented history of an object. The more provenance there is, the more reliable we can consider the estimation of authenticity. The third is material forensics testing—X-rays, ultraviolet or infrared analysis, dendrochronology, and similar techniques—which are very good at flagging anachronisms.
Now there is a fourth pillar, really only in recent years, which is AI image analysis. What firms like Art Recognition do is carefully and slowly train their AI in a closed data set to know everything there is to know about an artist in question. The AI is learning from all three traditional pillars. It’s being shown confirmed works by the artist, and also works we know are not by the artist, to teach it what not to be fooled by. That’s extremely important. If you’re asking an AI image analysis firm to test something, you need to ask very carefully what basis the AI is using to evaluate the work. It will not work if it’s open to the internet. Showing ChatGPT an artwork and asking, ‘Is this by Picasso?’ will not work because the internet is full of useless and false information as well as useful information. It has to be a closed data set, and you need as many reference images as possible. Some firms claim to use AI and have, I don’t know, ten images, and say they’ll test against those ten images. That’s not definitive.
The main issue now is that human connoisseurs are feeling a little nervous about AI because they’re afraid it’s going to push them aside—which, of course, it will not do. Human experts were similarly nervous when forensic testing first emerged. But AI image analysis is simply another tool. It’s taught by human scholars, based on information from human scholars, and still needs to be evaluated by human scholars. Ideally, any object should be subjected to all four pillars: the human eye, provenance research, material forensic testing, and AI image analysis. But because AI image analysis incorporates the other pillars and is relatively inexpensive—it can cost just thousands to test an object and only requires a high-resolution digital image—it’s becoming the easiest and most effective initial point of entry.
On forgery and value
In The Art of Forgery, you examine how convincing fakes are constructed. Where and how do forgeries still manage to deceive even top specialists today?
I wrote The Art of Forgery back in 2015, and I’m pleased to say it still remains the definitive book on the subject. I’m now releasing a new edition with Bloomsbury and had to add about 10,000 new words about the digital world.
The lessons, however, remain the same. Forgers are still active and still deceiving people in the exact same ways they traditionally did. The main issue is that we now have technology capable of exposing forgeries, but the question is whether people are going to use it.
If every artwork worth over $100,000 were subjected to material forensic testing and AI image analysis, we would have almost no forgeries in the high-end market. The problem is that many forgeries exist at the lower end of the market, where people don’t consider testing cost-effective. So the same systems remain in place for how forgers pass off their works and integrate them into the market. We simply now possess better tools to catch them.
On provenance risks
You’ve written about the ‘Provenance Trap.’ Where do specialists and collectors most often go wrong when assessing provenance in today’s market?
The trade looks for as much provenance as possible because more historical points of reference are generally considered better. Normally, that’s a good thing. But the question is whether those historical points of reference are definitively linked to the object in question.
What we sometimes see is forged provenance—or entirely invented provenance—mixed in among genuine references. It looks authentic, and people don’t think to double-check every detail because they assume the researcher has already done that work. I outline five variations of what I call the ‘Provenance Trap’ in the new edition of The Art of Forgery. It’s the most frequently used, sophisticated, and successful method by which forgers or their accomplices pass works into the market. It’s based on overenthusiasm and overreliance on provenance among researchers. Sometimes, if the provenance looks compelling enough, people won’t actually look at the object itself as carefully as they should.
That has led even fairly ridiculous-looking forgeries—like those by Mark Landis—to be accepted, at least briefly, because the story behind the work was compelling enough. Again, if these objects were subjected to forensic testing and AI image analysis, I think the percentage of successful forgeries entering the market would become extremely small.
On surprising cases
What is the most surprising case you’ve encountered in your research—whether a work believed to be genuine that turned out to be fake, or the reverse?
I love surprises. I have a new podcast called Is It? The Art Mystery Podcast, where I invite experts to choose a painting, and then we analyse it using AI image analysis before revealing the results live on the show. In just the first season, we’ve had four major discoveries that either upgraded or downgraded works in ways that contradicted scholarly consensus.
One forthcoming episode examines a painting owned by the National Gallery in the United States that they believed was not by Vermeer, while specialists at the Rijksmuseum believed it was. We tested it, and it turns out it actually is by Vermeer. What’s interesting is that the owning institution was being conservative. You would think they would have a vested interest in optimistic attribution, but in fact they really did have a Vermeer.
One of the funniest stories, though, involves the German forger Lothar Malskat, who forged medieval-style frescoes in a church after the Second World War. He did such a convincing job that nobody believed he was the forger. So he made the extraordinary move of suing himself in order to prove publicly that he had forged them.
He had inserted what are called ‘time bombs’ into the frescoes—intentional anachronisms designed to prove authorship if nobody believed him. Among them were a highly realistic image of Marlene Dietrich and a turkey, which could not have appeared in medieval Europe before the discovery of the Americas. He successfully proved he was the forger, meaning he simultaneously won and lost the case. It remains one of the funniest stories in art crime history.
On illicit art circulation
In 2026, what are the most common ways stolen artworks resurface in the market?
It’s probably still online auction sites. The dark web exists, of course, but not to the extent people imagine. Most illicit circulation happens through ordinary online platforms.
This is especially true for looted antiquities and non-unique objects like ancient coins or pre-Columbian figurines. If you browse online auction sites, it’s probable that much of what you see in terms of antiquities is either a forgery, a copy, or looted or stolen. Authentic objects with proper legal paperwork and ethical provenance are generally sold through galleries or major auction houses because sellers can obtain significantly higher prices there.
We’re not really seeing dramatic changes in the ways objects are stolen. What has changed is that more of the trade has moved online because sellers can now access the entire world as a clientele. Again, the technology to inhibit the sale of stolen objects or forgeries exists. It’s simply not being used as universally as it should.
Professor Noah Charney Website | ARCA | Podcast: Is it? – The Art Mystery Podcast
