Instant messaging (IM) applications, especially WhatsApp, have become ubiquitous in contemporary computer-mediated communication practices. IM data have the potential to constitute a rich source of research material for corpus linguistics and cultural analytics, owing to their similarities with face-to-face conversations as well as their private nature. In this work, we outline the creation process of a large curated dataset of WhatsApp messages in French. The paper covers the protocol for collecting these messages as well as the de-identification process for removing sensitive information liable to identify the users in these messages. The de-identified dataset will ultimately be made available to researchers on request.