Over the past few decades, English has become embedded as a key language for scholarly communication in many disciplines, but the approach of using a primary language for publication brings a number of disadvantages to individual scholars, to research, and to society. During this same period, the underlying approach used design and implement translation technologies has experienced a paradigm shift and now integrates AI-based techniques such as machine learning. While these translation tools hold promise for supporting a more multilingual scholarly communication eco-system, they cannot simply be adopted wholesale. Performance varies from one language to another, as well as from one discipline to another. In this workshop, we will explore how data-driven technologies work, where their strengths and limitations lie, and how we can interact with them in an informed and responsible way as we work towards the goal of increasing multilingualism in scholarly publishing.