This study investigates how artificial intelligence (AI) applications shape learner agency in the context of Legal English instruction. While AI has been widely integrated into language learning, its use in specialized domains such as Legal English remains underexplored. Legal English, being a domain-specific register characterized by precise terminology, complex syntax, and professional communicative conventions, requires learners not only to acquire linguistic knowledge but also to develop critical judgment and agency. Drawing on qualitative inquiry, this research examines how learners of Legal English exercise agency when engaging with AI-supported learning platforms. The study focused on postgraduate law students enrolled in a Legal English course that employed AI-driven tools such as machine translation, intelligent writing assistants, and automated feedback systems. Findings suggest that AI support both enables and constrains learner agency: while it empowers learners to independently draft, revise, and evaluate legal texts, it also creates dependencies that may weaken critical thinking if not mediated by pedagogical scaffolding. The study contributes to ongoing debates on the pedagogical role of AI in language teaching for specific purposes (LSP), highlighting that fostering learner agency requires balancing AI’s efficiency with opportunities for reflection, negotiation of meaning, and critical evaluation.