The development of efficient and safe nucleic acid delivery vectors remains an unmet need holding back translation of gene therapy approaches into bedside. Graphene oxide (GO) could help bypass such bottleneck thanks to its large surface area, versatile chemistry and biocompatibility, which could overall enhance transfection efficiency while abolishing some of the limitations linked to the use of viral vectors. Here, we aimed to assess the capacity of bare GO, without any further surface modification, to complex a short double-stranded nucleic acid of biological relevance (siRNA) and mediate its intracellular delivery. GO formed stable complexes with siRNA at 10:1, 20:1 and 50:1 GO:siRNA mass ratios. Complexation was further corroborated by atomistic molecular dynamics simulations. GO:siRNA complexes were promptly internalized in a primary mouse cell culture, as early as 4 h after exposure. At this time point, intracellular siRNA levels were comparable to those provided by a lipid-based transfection reagent that achieved significant gene silencing. Time-lapse tracking of internalized GO and siRNA evidenced a sharp decrease of intracellular siRNA from 4 to 12 h, while GO was sequestered in large vesicles, which may explain the lack of biological effect (i.e. gene silencing) achieved by GO:siRNA complexes. This study underlines the potential of non-surface modified GO flakes to act as 2D siRNA delivery platforms, without the need for cationic functionalization, but warrants further vector optimization to allow effective release of the nucleic acid and achieve efficient gene silencing.