A recently developed classification technique is proposed as a tool for e-mail filtering. Evolutionary support vector machines (ESVMs) consider the existence of separating hyperplanes between samples and determine the parameters of these decision functions through evolutionary algorithms (EAs). ESVMs can be regarded as the evolutionary counterpart of state-of-the-art support vector machines (SVMs). The novel paradigm is applied for solving the increasingly dangerous problem of spam e-mails, by automatically differentiating them from regular e-mails, solely by using their contents. Obtained results encourage further investigation.