As generative systems are absorbed into the linguistic infrastructure of multilingual states, the question of how their use is governed has acquired material consequences for citizens who depend on translation to reach public institutions. Yet the ethical governance of artificial intelligence and the study of translation policy have developed largely in isolation. AI-ethics scholarship has converged on a recognisable repertoire of principles—transparency, fairness, accountability, non-maleficence—while saying little about language; translation-policy scholarship theorises how states manage the linguistic dimension of public communication but has not been brought systematically to bear on AI-mediated communication. This paper joins the two. It asks how far the AI-ethics frameworks now shaping translation and communication systems converge across jurisdictions, and where they diverge in ways that matter for linguistic diversity.

The study reports a structured qualitative content analysis—the systematic coding of each document against an a priori codebook, with an intercoder-reliability check (κ = 0.86)—of 39 national AI policy and governance instruments (2020–2026), not legislation, from five deliberately contrasting jurisdictions: Qatar, the United Arab Emirates, Saudi Arabia, the European Union, and Singapore. Because the same coding scheme is applied uniformly across all five, the comparison rests on the coded provision rather than on any shared legal system. Drawing on González Núñez’s account of translation policy and Meylaerts’s model of translational regimes, and reading these against Mittelstadt’s critique of principle-based AI ethics, the analysis codes each document both for its ethical commitments and for the translational regime it presupposes, before mapping cross-jurisdictional patterns through matrix analysis.

The analysis reveals a shared principlist vocabulary overlaying markedly different translational regimes, and a recurrent under-elaboration of provisions for linguistic diversity and cultural sensitivity relative to other governance dimensions. The paper argues that principlism, transposed into multilingual settings without a translation-policy supplement, systematically under-specifies the linguistic conditions of ethical AI communication, and it identifies where that gap is most consequential for linguistically plural societies.

Acknowledgement: This work was supported by Qatar University Internal Collaborative Research Grant No. QUCG-CAS-26/27-921. The findings achieved herein are solely the responsibility of the author.