Large Language Models have already begun to affect the higher education landscape. However, there is currently a lack of work investigating how these models interface -- and possibly interfere -- with literacy development. Considering literacy is critical because student learning is only made possible through language. This paper considers implications for university students' literacy development by drawing on Lea and Street's academic literacies framework. I argue that LLMs pose different levels of risk for students' development of each aspect of literacy contained within the framework: study skills are least at risk, academic socialisation is most at risk, and academic literacies represent an intermediate risk. Implications for instructors include dedicated instructional time and support for students to engage with reading and writing practices without LLM support before they begin to incorporate them into their literacies repertoire. If students rely too heavily on LLMs initially, there is a danger they will not undergo the enculturation and cognitive development necessary for success at university.