Gene therapy for cancer treatment represents a promising approach that has shown selectivity and efficacy in experimental systems as well as clinical trials. Some major problems remain to be solved before this strategy becomes routinely adopted in the clinic, one of the main challenges being the improvement of gene delivery. Namely, the development of DNA vectors characterized by maximum efficiency and minimal toxicity will define the success of gene therapy and its chances of being accepted by public and clinicians. A number of issues need to be considered. The "magic" vector should be targeted, protected from degradation and immune attack, and safe for the recipient and the environment. Moreover, it should express the therapeutic gene for as long as required, in an appropriately regulated fashion. Vehicles such as retroviruses, adenoviruses and liposomes have been adopted in clinical studies, with varying results. New therapeutic modalities are also being explored in order to overcome the limitation of poor gene transfer and patient toxicity, including bacteria, adeno-associated and herpes simplex viruses, lentiviruses, cationic polymer-DNA complexes and electroporation. Some of the delivery systems tested in preclinical and clinical models are reviewed in this article, with particular attention to the targeting of the tumor environment.