<p>Artificial intelligence technology has posed significant challenges to legal translation. However, few scholars have analyzed the performance of artificial intelligence in the translation of terms. By constructing a corpus of the terms in the ChatGPT-generated text of the <i>Civil Code of the People’s Republic of China (Contract)</i>, the paper analyzes the performance of ChatGPT in the English translation of terms from the perspective of functional theories of translation. The result shows that ChatGPT does not perform well from the perspective of functional theories of translation on the whole. Firstly, by analyzing ChatGPT English translation of terms from the perspective of functionality, the paper finds that ChatGPT does not perform well in the realization of the referential function, appellative function and expressive function. However, in terms of the realization of phatic functions, ChatGPT performs well when translating the terms that convey the same concept in the common law system and the Chinese legal system, but ChatGPT does not perform well when translating the terms that convey different concepts in the common law system and the Chinese legal system. Secondly, by analyzing ChatGPT English translation of terms from the perspective of loyalty, the paper finds that ChatGPT does not perform well in the expression of intercultural concepts and the legislator’s communicative intention. The paper enlightens us that artificial intelligence-driven tools remain proficient in translating culturally overlapping terms while having difficulty in addressing the complexities of culture-specific legal expressions and conceptually divergent terms across different legal systems.</p>