Archimag : The specialized reference in digital information management
More than 57,000 objects are currently listed in Interpol’s database of stolen works of art. According to the international police organisation, this figure reflects a broader reality: trafficking in cultural property is both low-risk and highly profitable for criminal networks.
Recent high-profile cases illustrate the scale of the issue. Every day, somewhere in the world, cultural objects are stolen or looted to be sold illegally on the market. Over the past three decades, illicit trafficking in artworks has become a major global concern.
AI in the fight against art theft
In response, institutions are increasingly turning to AI. Interpol has developed the ID-Art application, which relies on advanced image recognition technology. The tool allows users - including police officers, customs agents, art dealers or private individuals - to photograph a suspicious artwork. The algorithm then compares the image against Interpol’s global database.
The result: potential matches can be identified almost instantly. When a similarity is detected, users are encouraged to submit a report for further investigation.
During its pilot phase, the application contributed to the identification of several stolen works, including statues in Italy and paintings in Amsterdam that had reappeared on online marketplaces.
Monitoring the art market - including the dark web
Other initiatives go even further. The Italian Carabinieri Command for the Protection of Cultural Heritage has developed an AI-powered platform called Swoads (Stolen Works of Art Detection System). Connected to Italy’s national database of cultural property, it currently tracks more than 1.3 million stolen items.
The system continuously scans online marketplaces, auction catalogues and even parts of the dark web. Combining image and text recognition, it identifies potential matches and generates alerts, which are then reviewed by specialised officers.
After several years of use, the platform is being enhanced with technologies such as big data, machine learning and blockchain. This allows it to combine automated data collection - from the web, deep web and social media - with operational intelligence gathered in the field.
The system is also being extended to detect looted or counterfeit objects listed by the International Council of Museums. To date, it has contributed to the recovery of more than 200 stolen items.
Also read: AI and scientific fraud: poison or cure?
A Dürer - with 82.2% probability
AI is not only used by law enforcement. It is also increasingly involved in art authentication.
In 2024, a private collector asked the Swiss company Art Recognition to analyse a portrait attributed to Albrecht Dürer. The result: a probability of 82.2% that the artwork is genuine.
To reach this conclusion, the system was trained on a dataset of 144 authenticated works by Dürer, along with an equivalent number of forgeries, imitations and synthetic images. The AI learned to recognise the artist’s distinctive style based on visual patterns.
Interestingly, the system also analysed a version of the same work held by the British Museum - long considered authentic - and assigned it a slightly lower probability score.
As the number of forgeries increases, such tools are gaining traction among collectors and cultural institutions. Founded in 2019, Art Recognition reports more than 500 authentication assessments and over 100 clients.
Analysing Raphael’s brushstrokes
In the United Kingdom, researchers at the University of Bradford have used AI to revisit the attribution of the painting Madonna della Rosa, traditionally attributed to Raphael.
Their analysis suggests that the face of Saint Joseph, in the upper left corner, may have been painted by another artist.
This conclusion is based on the analysis of brushstrokes, colour palettes and shading - details that AI can process more quickly and systematically than human experts when trained on large datasets.
However, researchers stress that AI is not intended to replace art historians. Authenticating a work involves multiple dimensions, including provenance, materials and conservation history.
AI should therefore be seen as an additional tool, supporting - rather than replacing - expert analysis.
