Large language models based on machine-learning technologies are reshaping linguistic contexts and understandings of language. We explore these reconfigurations by investigating discursive positionings of traditional institutional guardians of power in language in response to these changes. Focusing on the discourse of the Real Academia Española (RAE), we show how RAE’s social functions, ways of asserting authority, and the nature, function, and rightful ownership of RAE’s standard language have been reimagined. Crucially, RAE presents itself as a professional soft power that protects the rights of Spanish speakers. Drawing on tropes of authenticity and endangerment, it conceptualises language generated by machine-learning technologies as inauthentic and as destroying the authentic Spanish of human Spanish speakers. We argue that these discourses are indexical of a power struggle where the role of traditional language norming institutions is reshaped in the face of sociotechnical innovations that are in the hands of global commercial companies. (Standard language, AI technology, language academies, authority in language, big tech, Real Academia Española)*