AI-assisted translation: trust, authority, and human–AI mediation

The rapid expansion of AI-assisted translation raises pressing questions about how intercultural understanding can be achieved when humans and automated systems collaborate in linguistic mediation. Drawing on recent work in translation studies and intercultural communication, this contribution offers a conceptual synthesis of how AI-generated translation, including both neural machine translation and more recent large language model–based systems, reshapes the communicative, epistemic, and ethical conditions of translation. While contemporary AI systems can produce fluent output in high-resource language pairs, they are less efficient in contexts where cultural representation and linguistic diversity are most at stake. This can mask distortions and reinforce epistemic asymmetries across languages and cultures. Moving beyond polarized narratives of automation, the article argues that translation remains anchored in automation-resistant competences. These competences function not only as professional skills but also as conditions for trust and authority in AI-assisted workflows. Human translators are therefore understood not merely as post-editors but as epistemic mediators who exercise contextual judgement and safeguard communicative legitimacy. These dynamics are examined across professional practice and educational settings, including translator training and foreign-language programs where translation is used pedagogically to cultivate mediation competence. The discussion concludes that AI does not eliminate mediation in translation but redistributes and intensifies it. It makes human interpretive agency more central, not less, in an increasingly automated communicative landscape.