Cancer is most commonly viewed as resulting from somatic mutations enhancing proliferation and invasion. Some hypotheses further propose that these new capacities reveal a breakdown of multicellularity allowing cancer cells to escape proliferation and cooperation control mechanisms that were implemented during evolution of multicellularity. Here we critically review one such hypothesis, named "atavism," which puts forward the idea that cancer results from the re-expression of normally repressed genes forming a program, or toolbox, inherited from unicellular or simple multicellular ancestors. This hypothesis places cancer in an interesting evolutionary perspective that has not been widely explored and deserves attention. Thinking about cancer within an evolutionary framework, especially the major transitions to multicellularity, offers particularly promising perspectives. It is therefore of the utmost important to analyze why one approach that tries to achieve this aim, the atavism hypothesis, has not so far emerged as a major theory on cancer. We outline the features of the atavism hypothesis that, would benefit from clarification and, if possible, unification.