AI helps uncover stolen 'Saint Agnes' painting 20 years on despite repainting
The white lamb had vanished from the saint's arms, and the halo and palm branch had been painted out, so the picture looked like a portrait of a woman holding a scroll. The changes were enough to alter the work's identity for an ordinary visitor, but they did not fool an AI-based visual comparison system, which alerted Italian police that the painting might be stolen.
The story began in October 2024, when officers from the cultural heritage protection unit toured an antiques fair in the city of Parma. They photographed the works on display and fed the images into a specialised system that compares them with records of stolen items. Within hours, an alert came up for a painting of a woman in red and aquamarine holding a scroll, and investigators returned the next day to seize it for examination.
The examination identified the work as a 17th-century painting of Saint Agnes that was stolen in 2004 from the Virgo Potens church in Sulmona, in the Abruzzo region. Parts of it had been repainted to hide its religious symbols: the lamb became a scroll and the attributes by which the saint is usually recognised were removed. The painting was returned in 2025 to the diocese of Sulmona-Valva.
Visual recognition systems work by comparing the features of the image itself, such as the distribution of colours, shapes and composition, rather than relying on the artist's name, the title or a written description. That makes them useful when an object is shown under a different name or has had some details altered. An automated alert is not proof of identity, however; it is a lead worth following, and final verification remains the job of experts.
Al Majlis Analysis: This case offers a balanced picture of what artificial intelligence can do, without the hype. The machine did not solve the mystery on its own. It shortened a screening process that could have taken years and then handed the task to investigators and specialists. As a growing share of the art and antiques trade moves to online auctions and listings, where pieces appear and disappear quickly, the ability to compare millions of images becomes a decisive tool for protecting heritage. Yet the technology depends entirely on good reference photos and accurate records of stolen works. That is the key message for museums, places of worship and collectors everywhere, including in our region: documenting collections with photos and data today is what makes recovering them possible tomorrow.