ANCHISE Talks 1 - What can AI do to combat the illicit trafficking of cultural heritage?

ANCHISE Talks is the webinar series of the ANCHISE project, which explores the fight against the illicit trafficking of cultural goods and its innovative solutions. The round of conversations has been opened by Sorin Hermon of the Cyprus Institute, who delves into possibilities and challenges around Artificial Intelligence in enhancing the fight against illicit trafficking of cultural goods. Taking place on January 25, 2024, the webinar saw the participation of more than 80 professionals, researchers and students.

Main takeaways from the webinar:

The rise of AI and Web 3.0 offers new opportunities for protecting cultural heritage. Web 3.0, also called the Semantic Web, aims to create a decentralized Internet environment focused on data comprehension by machines. AI and machine learning are essential for selecting and analyzing vast datasets, enabling new applications in combating illicit trafficking.

Access to large, high-quality datasets is crucial for training effective AI models. Existing databases on stolen objects, such as those of Interpol, the Italian Carabinieri, and the FBI, are limited in terms of size and descriptive richness. Moreover, they often operate in isolation, hindering the development of efficient ML algorithms. Defining precise behavioral models is essential for identifying and authenticating illicit objects. These models must describe the typical patterns of illicit trafficking, from illegal excavation to sales on the black market or in auction houses. Collaboration between experts in the humanities, social sciences, heritage management institutions, and law enforcement is essential to provide the knowledge needed to create these models. Data quality is as important as quantity. The FAIR principle (Findable, Accessible, Interoperable, and Reusable) is a first step in ensuring data quality, but it is also crucial to consider relevance, reliability, reproducibility, and traceability of data. The semantic representation of data provenance is essential for ensuring transparency and reliability. Detailed descriptions of analysis methods, protocols, environmental conditions, and software used are necessary to assess data quality and allow reproducibility of results. [14, 15]

The concept of a digital twin for heritage (Heritage Digital Twin) offers a promising framework for organizing and visualizing complex data. The digital twin integrates all available information on a heritage asset, including its 3D description, history, materiality, conservation status, and provenance. Harmonizing terminologies, linguistic translation, and highlighting unique features of objects are essential for creating rich, interoperable data ecosystems. Large-scale European initiatives such as the European Collaborative Cloud for Heritage Data and the European Data Space for Cultural Heritage contribute to this effort.

According Dr. Hermon, AI offers considerable potential for combating illicit trafficking of cultural goods, but its success depends on the availability of high-quality data, precise behavioral models, and close collaboration between various stakeholders. The webinar highlights the need to invest in research, innovation, and the development of robust data ecosystems to effectively protect cultural heritage.