FOR IMMEDIATE RELEASE

New York, July 24, 2025 — Until now, AI-based authentication relied primarily on scanning hundreds of high-resolution images of confirmed authentic paintings in order to "educate" the system before submitting the questionable artwork for analysis. This method presents numerous challenges.

The most significant limitation is that approximately 300 authentic paintings must be scanned to properly train the AI system. In practice, this is often impossible. For example, only 18 universally accepted paintings by Leonardo da Vinci are known to exist, and only about 150 by Frida Kahlo. Even in the case of Rembrandt, the number of works firmly attributed to him has dropped to barely 300.

An additional challenge lies in the logistics of collecting 300 high-resolution images, which are typically scattered across museums, institutions, and private collections worldwide. Coordinating access to such a dataset is not only difficult and time-consuming, but also extremely costly — often requiring years of effort.

For these reasons, the traditional scan-train-analyze AI model has proven to be impractical, costly, and in many cases, simply unfeasible.

In response, Art Certification Experts has developed a second-generation AI system that eliminates the need to amass hundreds of known examples. Instead, the new approach focuses on extracting a wide array of specific, individually significant metrics directly from the questioned artwork. These include quantifiable elements such as brushstroke analysis, pigment composition, surface aging, and signature structure.

By using rich, granular data that can be rapidly extracted and assessed, the second-generation system provides a faster, more scalable, and more flexible way to leverage AI in the authentication process — especially in cases where few confirmed works exist.

This cutting-edge system will be fully operational and available to clients beginning September 2025.