Argument mining is a field concerned with building models that can automatically process argumentative text to find and relate the premises and claims contained in the text. While many models and papers have been published for argument mining with English text, very few have been published for other languages. This paper applies a cross-lingual approach to build an argument mining model for the Arabic language through translating an English corpus and projecting the labels. This corpus is then used to train an end-to-end BERT-based model. Although this approach has shown promising results in some European languages, the results of this paper show that this approach is less effective for Arabic (F1 scores of about 44%). On the other hand, classifying each sentence discretely using BERT-based models yields strong results (F1 scores of about 90%); while it still underperforms for argument component classification, which suggests that more complex models might be necessary. © 2025 IEEE.