75

Recent advances in artificial intelligence (ai) based on machine-learning methods have led to the proliferation of talking machines in work environments. A major objective since the very beginnings of ai, the development of machines’ ability to talk, has long raised concerns and questions about what talking means. This article updates this reflection by analyzing three automated language practices: the dubbing in real time of Indian telephone operators in order to mask their accent, voice assistance at the workstation in industry, and banking assistance via a chatbot. The authors show that behind the promises of performance, efficiency, productivity, and improved working conditions, these talking machines feed mechanisms of cultural, organizational, and male domination. Finally, they question the relevance of the European Union’s regulations, which are being developed to align these machines with fundamental rights and principles.