If it ain’t Baroque, don’t fix it — unless, of course, AI tells you that the so-called copy on your wall is actually a Caravaggio.
Earlier this week, reports revealed that AI analysis has overturned decades of skepticism, indicating that a long-dismissed “copy” of Caravaggio’s The Lute Player may actually be an authentic work by the Baroque master.
From copy to contender
For decades, the work was dismissed — Sotheby’s sold it in 1969 as a copy “after Caravaggio” for just £750, and in 2001 as a “circle of Caravaggio” for about £71,000.
Now, artificial intelligence has flipped the tables, assigning an 85.7% probability that the painting is actually by the Baroque master himself. Experts caution that AI cannot provide absolute proof, but in this case, its verdict is considered unusually strong.
Caravaggio, or Michelangelo Merisi da Caravaggio, is one of the most revered artists in Western history, active at the turn of the 17th century. Only a few dozen of his works survive, so each attribution is significant.
When a previously unknown Caravaggio surfaced in 2019, it was valued at nearly £96 million.
The Badminton House painting
The rediscovered Lute Player once hung in Badminton House in Gloucestershire. Art historian Clovis Whitfield purchased it in 2001, recognizing details that matched a 1642 biography of Caravaggio by Giovanni Baglione.
“Baglione mentions minutely observed details such as the reflection on dew drops on the flowers,” Whitfield said, according to The Guardian.
This is one of three known versions. The second, considered undisputed, hangs in the Hermitage Museum in Russia. The third — nicknamed the Wildenstein version — replaces the young man with a woman and was on long-term loan to the Met from 1990 until 2023.
For years, the Wildenstein version received recognition as the “real” Caravaggio, while the Badminton canvas was quietly written off as the understudy.
AI tips the scales
The new analysis came from Art Recognition, a Swiss company working with the University of Liverpool.
“Everything over 80% is very high,” Dr. Carina Popovici, the company’s head, told The Guardian. Their AI found a “strong match” between the Badminton painting and known Caravaggios.
Even more surprising, the same system concluded that the Wildenstein version was “not an authentic work.” Lute expert David Van Edwards supported the finding, noting that the Wildenstein instrument was riddled with inaccuracies, unlike the faithful depictions in the Badminton and Hermitage versions.
“The AI result knocks Mr. Christiansen off his perch,” Whitfield said, referring to Keith Christiansen, the Met’s former head of European paintings, who long dismissed the Badminton version.
A growing chorus
Others are rallying to the Badminton side. William Audland KC, a barrister writing a book on the painting, is one such defender.
“Taking all the evidence into account, it seems to me that a manifest injustice is being done by any scholar who suggests that the Wildenstein version is autograph and the Badminton version is a poor copy,” he said. “A holistic view… has now been corroborated by AI analysis, which is objective, unlike the subjective opinions of scholars. The Badminton version is an astonishing painting. It takes your breath away when you see it.”
In an interview with Artnet, Popovici said that “this case highlights again the rigidity of the current market and the way in which strongly held expert views can hinder progress.” She expressed her hope that using AI “can contribute to addressing such situations with greater openness and less conflict.”
The painting now resides in London, where Whitfield hopes it will one day enter a public collection.
The Caravaggio finding highlights one side of AI’s growing role in art: revealing what’s real. On the other side, tools like Midjourney and DALL·E are focused on creating the unreal — new visions, new styles, and new debates about originality.
The post AI Declares £71,000 ‘Copy’ a Genuine Caravaggio appeared first on eWEEK.
No Responses