One function ascribed to apoptosis is the suicidal destruction of potentially harmful cells, such as cancerous cells. Hence, their growth depends on evasion of apoptosis, which is considered as one of the hallmarks of cancer. Apoptosis is ultimately carried out by the sequential activation of initiator and executioner caspases, which constitute a family of intracellular proteases involved in dismantling the cell in an ordered fashion. In cancer, therefore, one would anticipate caspases to be frequently rendered inactive, either by gene silencing or by somatic mutations. From clinical data, however, there is little evidence that caspase genes are impaired in cancer. Executioner caspases have only rarely been found mutated or silenced, and also initiator caspases are only affected in particular types of cancer. There is experimental evidence from transgenic mice that certain initiator caspases, such as caspase-8 and -2, might act as tumor suppressors. Loss of the initiator caspase of the intrinsic apoptotic pathway, caspase-9, however, did not promote cellular transformation. These data seem to question a general tumor-suppressive role of caspases. We discuss several possible ways how tumor cells might evade the need for alterations of caspase genes. First, alternative splicing in tumor cells might generate caspase variants that counteract apoptosis. Second, in tumor cells caspases might be kept in check by cellular caspase inhibitors such as c-FLIP or XIAP. Third, pathways upstream of caspase activation might be disrupted in tumor cells. Finally, caspase-independent cell death mechanisms might abrogate the selection pressure for caspase inactivation during tumor development. These scenarios, however, are hardly compatible with the considerable frequency of spontaneous apoptosis occurring in several cancer types. Therefore, alternative concepts might come into play, such as compensatory proliferation. Herein, apoptosis and/or non-apoptotic functions of caspases may even promote tumor development. Moreover, experimental evidence suggests that caspases might play non-apoptotic roles in processes that are crucial for tumorigenesis, such as cell proliferation, migration, or invasion. We thus propose a model wherein caspases are preserved in tumor cells due to their functional contributions to development and progression of tumors.