Anthony Van Dyck: “The portrait of Don Felipe de Guzmán”
The evolution of art authentication has come a long way since traditional connoisseurship. Complex and multifaceted, it is a process that often encounters challenges and controversies. While the expert eye of the connoisseur remains a solid pillar, new technologies – ranging from scientific methods introduced in the 19th century, like spectroscopy and X-ray analysis, to recent advancements in AI – are beginning to reshape the field. Yet, despite these advances, the integration of AI in art authentication has been met with resistance. This distrust still lingers among many art experts, and Niels Büttner – Professor of Medieval and Modern Art History at the State Academy of Fine Arts in Stuttgart – was one of them. He expressed his skepticism for AI’s role in art authentication in an article in 2023, criticizing Art Recognition’s technology. But, this initially turbulent relationship has transformed into a very productive collaboration, resulting in a joint article.
An illustrative example of how two different approaches can complement one another. This interdisciplinary collaboration has blended art historical expertise with cutting-edge AI technology to authenticate a painting attributed to the Flemish master Anthony van Dyck. The painting, depicted in Fig. 1, represents the portrait of Don Diego Messía Felipe de Guzmán, Marqués de Leganés. This artwork has been the subject of much debate among scholars, who consider it to be a creation of Van Dyck’s workshop or one of his students, rather than by the master himself. The original version of the painting, actually created by Van Dyck, is exhibited at the Tokyo Museum (Fig. 2), while a second version is housed at the Fundación Santander in Madrid (Fig. 3).
A confirmation. For this AI authentication project, Art Recognition created a dataset of training images validated by Niels Büttner himself. Furthermore, Büttner examined the painting in question and declared that it was not autograph by the Flemish master, thus agreeing with previous scholars’ statements. The artificial intelligence model from Art Recognition also concluded that the painting is not authentic, with a 79% probability. By combining the insights of traditional connoisseurship with the objectivity of AI, this project exemplifies the potential of interdisciplinary collaboration in solving complex challenges of art authentication. The project not only resolved a long-standing debate but also paved the way for future collaborations that merge human connoisseurship and modern innovation.
