The current paradigm states that cancer progression is caused by random independent mutations, each selected for its survival advantages. The accelerated rates of phenotypic changes, the pleiotropic effect of several genes involved in progression--which need not be necessarily mutated for inducing the observed changes in cancer cell behaviour--lead us to propose an alternative hypothesis. Malignant progression might be a result of the unveiling of a cell-survival program, induced by various aggressions in the same way as the SOS system is induced and regulated in bacteria. This hypothesis depends on the homology between several genes involved in cancer progression (such as bcl2, mdm2, the mismatch repair genes, the heat shock protein genes, the pleiotropic resistance genes, the telomerase gene ...) and several genes involved in the survival of prokaryotes and eukaryotes under stress. The development of multicellular organisms could not take place without the building of a control program, exemplified by the so-called anti-oncogenes. However, this control program had to integrate some weaknesses, in order to allow for embryogenesis, growth, and wound healing. These weaknesses, neutral from an evolutionary point of view--since most cancers are sporadic and kill their hosts long after the birth of the offspring--are exploited by the survival program of individual cells, inherited from the genome of prokaryotes and unicellular eukaryotes, and repressed but not suppressed in animals. If this theory is true, it is probable that (i) no anti-oncogenes will be found in unicellular organisms, (ii) the sensitivity to mutations will be higher in genes involved in proliferation and in anti-oncogenes such as p53 and Rb, than in genes not involved in the cancer process, (iii) a process of transfer of genetic information exists in cancer cells as it exists in bacteria. The identification of the genes governing the survival program could lead to new therapeutic approaches.