Whose voices count? Multilingualism, AI policy, and equity at Hispanic-Serving institutions

Abstract

Universities have been rapidly developing generative AI policies to govern teaching, learning, research, and administration. Although these policies are typically framed around innovation, academic integrity, and risk management, they also shape language practices and participation in multilingual institutional contexts. Drawing on language policy scholarship that conceptualizes policy as discourse-in-action, this study examines AI policy discourse at two Hispanic-Serving Institutions (HSIs) in the United States. Using critical discourse analysis of AI guidance documents, syllabus templates, faculty guidelines, task force materials, and administrative communications, we identify two interrelated patterns across both institutions. First, AI governance constructs an implicitly English-dominant user while leaving multilingual practices and perspectives largely unaddressed. Second, responsible AI use is framed through procedural compliance and individualized responsibility, creating uneven terrain for multilingual users to navigate. Together, these patterns suggest that institutional AI policies function as de facto language policies that reproduce monolingual assumptions through both what they explicitly address and what they leave unsaid. Read against each institution’s broader commitments to multilingual communities, these findings indicate that AI governance and institutional linguistic priorities have not yet been fully integrated. We conclude by proposing a five-dimension continuum for more linguistically responsive and participatory AI governance across domains of governance, language orientation, ethics, responsibility, and innovation.