Beyond the hype: Why AI still cannot replace human translators and interpreters in high-stakes contexts

Abstract

Recent advances in artificial intelligence (AI) have significantly improved machine translation and real-time interpreting, fostering optimism about faster and more accessible cross-linguistic communication. However, this feature article argues that such developments pose critical limitations when applied to high-stakes public service domains such as legal, medical, immigration, and asylum contexts. In those settings, communicative failure can have profound consequences. Drawing on research from interpreting studies, sociolinguistics, and AI ethics, this study explains how language in these environments is not merely informational: it is also performative, culturally embedded, and deeply influenced by power relations, trust, and accountability. While AI systems may achieve high accuracy in non-critical interactions, they struggle with ambiguity, cultural nuance, minority languages, emotional distress, and unpredictable discourse conditions. Moreover, a range of issues including bias, confidentiality, ethical responsibility, and legal liability raise serious concerns regarding the substitution of human interpreters by AI.