Italy’s Carabinieri say their AI system, SWOADS, has helped identify and recover stolen art and antiquities worth more than $200 million. The system matched a disguised 17th-century painting of Saint Agnese to a work stolen from a church near L’Aquila two decades earlier, showing how image analysis can surface leads that are difficult to spot by eye.
A painting altered to hide its identity
The match began in October 2024, when officers from the Carabinieri Command for the Protection of Cultural Heritage visited Mercanteinfiera, an antiques fair in Parma. They photographed objects on display and uploaded the images to the police database. SWOADS flagged a painting of a young woman holding a scroll as a possible match for a missing work.
The original was a 17th-century oil painting of Saint Agnese, taken from Virgo Potens, a small church in Sulmona, in the Abruzzo region. In the stolen painting, the lamb the saint had been holding was covered over and replaced with a scroll bearing Latin text. That alteration made the work look different from the police image of the original.
SWOADS compared visual features including colour shades and rough silhouettes, identifying similarities despite the change. The painting was returned to the church. The people responsible for the theft have not been apprehended.
How SWOADS supports investigations
The Carabinieri Command for the Protection of Cultural Heritage, Italy’s art crime squad, has used the platform since 2023. IT engineer Salvatore Rapicavoli developed SWOADS, short for Stolen Works of Art Detection System, and joined the art crime team in 2023.
The system searches images against police records to flag possible matches. That matters when an object has been altered, photographed from a different angle or presented in a setting far removed from its original home. AI can narrow the search, but investigators still need to assess a match and follow up on the lead.
The Saint Agnese case shows the distinction: the software identified a visual resemblance, while police officers had to examine the painting and connect it to a specific theft and church.
Thousands of objects, from statues to coins
SWOADS has helped recover more than 300 large paintings and statues, alongside more than 165,000 smaller objects, including coins, letters and medals. Together, the recovered works and objects are valued at more than $200 million.
The range matters for a database built around images. Large paintings may offer broad visual patterns; smaller objects require recognition across details and forms that can be harder to distinguish. The system’s role is to compare those images and direct attention to potential matches, not to replace the investigation that establishes an object’s history.
What happens next
The Saint Agnese painting is back at Virgo Potens, but its recovery has not resolved the theft: the thieves remain at large. Investigators must still pursue the people behind the crime, a step that image recognition alone cannot complete.
The next concrete question in this case is whether police can identify and apprehend those responsible. For SWOADS, the process continues through the same cycle: compare images, examine flagged objects and use the resulting leads to support further investigation.
Sources
- artnews.com — “Italian Police Used AI to Identify $200 M. Worth of Stolen Art and Antiquities”
- Kunal S — “When cutting-edge technology meets the art world, centuries-old thefts can finally be unraveled. Italian police have turned to artificial intelligence to recove”
- artnews.com — “ARTnews.com – The Leading Source for Art News & Art Event Coverage”
- gadgetsnow.indiatimes.com — “How Italian Police Are Using AI to Track Down Stolen Art, One Disguised Painting at a Time”
- BBC — “New York returns $19m worth of stolen art to Italy”