Italian Police Use AI Art Theft Software to Find Loot
Italian police recovered a 17th-century painting stolen two decades ago after deploying new AI art theft software at an October 2024 antiquities fair in Parma. The recovery of the heavily disguised masterpiece demonstrates how law enforcement is increasingly automating the hunt for plundered cultural heritage.
The Stolen Works of Art Detection System (Swoads) flagged a painting of a woman in an aquamarine and red dress holding a half-unrolled scroll. Investigators seized the artwork the following day. It turned out to be Saint Agnese, stolen from the Virgo Potens church in Sulmona on August 11, 2004. Thieves had painted over the saint’s halo, palm branch, and lamb, replacing the animal with a scroll to evade traditional keyword searches.
Because the theft occurred more than 20 years ago, Italian law prevented prosecution. The Naples antiquities dealer who possessed the painting likely inherited it through a family business and knew little about its actual provenance.
Conceived in 2020 and officially introduced in November 2023, Swoads looks past superficial alterations. It matches the underlying visual fingerprints of an artwork, such as specific color shade combinations and rough silhouettes. This structural matching allows the system to identify pieces of a painting that were chopped up and sold separately, a common fate for large altarpieces.
The technology operates continuously, crawling e-commerce sites, online catalogues, social media, and the dark web. It has already helped Italian authorities recover more than 300 large works and 165,000 smaller items, totaling over €203 million in value. In early 2026, the unit recovered Filippo Palizzi’s Riposo della Pastorella, a painting stolen from a Turin modern art gallery in June 1972, after a routine AI sweep spotted it on an auction house website.
