Rapidly developing technological advances have raised new questions about what makes us uniquely human. As data and generative AI become more powerful, what does it mean to learn, teach, create, make meaning, and express ourselves, even as machines are trained to take care of these tasks for us? With youth, and in the context of literacy and media education, we embrace this moment to broaden our social imaginations. Our collaboration with journalists ages 14-25 from 2019 to 2023 has yielded a corpus of over 30 multimodal compositions constructed with and/or about AI reaching audiences in the millions. On the basis of these youth texts -- produced within our participatory research at YR Media, a national STEAM learning center and platform for emerging BIPOC content creators -- we developed the conceptual framework presented here: "Humanizing Data Expression" (HDE). The key role of expression in HDE distinguishes the human from the machine through the lens of storytelling. Analysis of this corpus (podcasts, web-based interactives, videos, radio features, online posts, social media assets) revealed four literacy practices of YR Media authors as they made sense of AI: (1) "contextualize": try out AI-powered features, reveal how it works; (2) "unveil authorship": introduce AI creators and processes; (3) "grapple": explore tensions and paradoxes; (4) "play": hack, mess with, outsmart, exaggerate AI. From these insights, we end with implications of HDE as a framework for learning and teaching AI literacy, including its potential for critically transforming data literacy practice and pedagogy across schools, teaching, and teacher education.