Machine Translation of Audiovisual Media and Epistemic Access: Ethical Questions in AI Video Localization

This paper examines AI-mediated video translation as a question in applied ethics, philosophy of language, and social epistemology. Video localization systems now combine speech recognition, subtitle translation, AI dubbing, speaker-aware voice generation, and lip-sync adaptation in a single workflow. These systems promise broader access to educational, professional, and cultural video content, but they also introduce risks involving mistranslation, loss of pragmatic context, voice identity, consent, and audience transparency. The paper argues that audiovisual machine translation should not be assessed only by linguistic accuracy. It should also be evaluated by how it preserves speaker intention, communicates uncertainty, supports accessibility, and avoids misleading viewers about the origin and status of translated speech. A short case discussion of contemporary AI video translation workflows is used to ground the argument without treating any tool as a product recommendation.