Therapeutic RNA interference (RNAi) has emerged as a promising approach for the treatment of many incurable diseases, including cancer, infectious disease or neurodegenerative disorders. Demonstration of efficacy and safety in animal models is necessary before planning human application. Our group and others have previously shown the potential of this approach for the dominantly inherited neurological disease DYT1 dystonia by achieving potent short-hairpin RNA (shRNA)-mediated silencing of the disease protein, torsinA, in cultured cells. To establish the feasibility of this approach in vivo, we pursued viral delivery of shRNA in two different mouse models. Surprisingly, intrastriatal injections of adeno-associated virus serotype 2/1 (AAV2/1) vectors expressing different shRNAs, whether targeting torsinA expression or mismatched controls, resulted in significant toxicity with progressive weight loss, motor dysfunction and animal demise. Histological analysis showed shRNA-induced neurodegeneration. Toxicity was not observed in animals that received control AAV2/1 encoding no shRNA, and was independent of genotype, occurring in both DYT1 and wild-type animals. Interestingly, the different genetic background of both mouse models influenced toxicity, being earlier and more severe in 129/SvEv than in C57BL/6 mice. In conclusion, our studies demonstrate that expression of shRNA in the mammalian brain can lead to lethal toxicity. Furthermore, the genetic background of rodents modifies their sensitivity to this form of toxicity, a factor that should be taken into consideration in the design of preclinical therapeutic RNAi trials.