The usage of artificial intelligence in legal translation: Bridging the gap between law and language

With the rise of Artificial Intelligence (AI) in the translation industry, translators are increasingly relying on AI-powered tools in their translation tasks to accelerate the translation process and improve the quality of their output. Undoubtedly, legal documents demand utmost precision, and while AI tools promise enhanced efficiency and consistency, their ability to fully grasp the intricate cultural and legal nuances that human translators master remains limited. In this context, the present research conducts a comparative analysis of Arabic-English legal translations produced by 27 university translation students and two AI-powered tools (ChatGPT and DeepSeek), examining three key dimensions of translation quality: accuracy, naturalness, and error typology, including lexical, syntactic, and pragmatic errors. The dataset comprises two authentic Arabic legal texts: one special power of attorney and one judicial pleading. They were selected to represent distinct legal genres and levels of linguistic and cultural complexity. We found that while AI excels in technical terminology and structural coherence, it often stumbles with culturally rich expressions and subtle contextual meanings. Human translators, despite occasional terminological inaccuracies, demonstrate a superior understanding of these critical cultural and legal elements. In conclusion, the study supports a collaborative approach that integrates the efficiency and accuracy of AI with the cultural awareness and contextual expertise of human translators. By adopting such a hybrid model, it becomes possible to bridge the gap between law and language, ultimately ensuring the production of high-quality legal translations in an increasingly interconnected world. The current study offers empirical evidence on the strengths and limitations of AI translation systems and human translation in the Arabic–English legal context, informing best practices for quality legal translation. The findings contribute empirical evidence to the growing body of research on AI-assisted legal translation, highlighting both the potential and limitations of large language models in Arabic-English legal contexts. • Translation students mistranslated key legal terms, relying on overly general or literal wording. • The human translation showed awkward sentence structures influenced by Arabic syntax. • ChatGPT’s English translation used literal structures but contained no major semantic errors. • DeepSeek’s literal rendering of honorifics produced culturally inappropriate language for English legal contexts.