Exploratory use of artificial intelligence (AI) has led to innovations across many fields in recent years. In view of these advances within high energy physics (HEP) and more broadly, we will (i) procure, develop and share benchmark datasets that adhere to the findable, accessible, interoperable, and reusable (FAIR) principles; we will (ii) demonstrate how to use FAIR benchmark datasets and share these AI models; we will (iii) lead community activities to define and implement FAIR principles for models and best practices for sharing HEP data and AI models, and we will (iv) demonstrate how the use of accelerated computing, scientific visualization, and domain-inspired training methodologies can lead to new insights into the interplay between data and models, and to the robustness of models to changes in architecture, hyperparmeter tuning, and to test data sets. These FAIR data sets will include metadata, provenance, and annotations to facilitate their use and define their scope of applicability. The FAIR data and AI models produced by this project will significantly reduce duplication of effort within the field of HEP, thus optimizing the use of DOE computational resources and facilities through innovative AI research. These activities will further DOE's objectives towards the construction of a theoretical framework that makes the best use of AI in science and engineering in preparation for the entry of HEP into the next generation of experiments, where the data volume will be measured in exabytes.
