Machine translation, which refers to the utilization of computational programs for conveying a stretch of words from one language into another, is quickly gaining ground, taking over the whole realm of translation and thus overthrowing human traditional translation. The paper principally aims to address the topic by: (1) presenting a comprehensive account of dysphemism in terms of its definitions, motivations, major types and functions. (2) Defining and analyzing the strategies employed in translating dysphemistic expressions to show whether the renderings arrived at by the machine are appropriate or otherwise. To achieve the previously-mentioned aims, it is hypothesized that (1) English and Arabic employ a variety of dysphemistic expressions. (2) Translating dysphemistic expressions poses unsolved problems for the machine as it does not possess the proper systematized inputs that enable it to efficiently process the pejorative data and adequately transfer the pragmatic content of the SL expression(s). To test the validity of these hypotheses, many authentic English translation examples are selected to be translated by the machine, a translation model is adopted to subtly apprehend dysphemism, data analysis is conducted, and new renderings are proposed wherever the machine renderings are found inappropriate. The paper ends with some conclusions and recommendations that help translators know the whereabouts in which the machine might fail to come up with sound judgment of the text and fall short of putting forth proper translation. The human translator should always be on the alert and watch out for any pitfalls of the kind and correctly edit them manually.