Machine translation (MT) platforms have gained increasing attention in the educational linguistics community. The current article extends past research on instructor beliefs about MT by way of an ecological theoretical framework. The study reports on a large-scale survey (n=165) of FL university-level instructors in the U.S. Findings indicate strong lines being drawn around acceptable MT use (e.g., in relation to text length and skill, policies), an acknowledgement of widespread student use driven by diverse motivations, and the Janus-faced nature of MT's potential threat to the profession. These findings reveal several salient tensions in how MT mediates relationships in language education (e.g., constructions of students, the nature of language and language learning, goals of the profession) that shed new light on the impact of MT technologies on the field. Implications for future research and the development of pedagogical practices anchored in digital literacies conclude the piece.