Restricting the cytotoxicity of anticancer agents by targeting receptors exclusively expressed on tumor cells is critical when treating infiltrative brain tumors such as glioblastoma multiforme (GBM). GBMs express an IL-13 receptor (IL13Rα2) that differs from the physiological IL4R/IL13R receptor. We developed a regulatable adenoviral vector (Ad.mhIL-4.TRE.mhIL-13-PE) encoding a mutated human IL-13 fused to Pseudomonas exotoxin (mhIL-13-PE) that specifically binds to IL13Rα2 to provide sustained expression, effective anti-GBM cytotoxicity, and minimal neurotoxicity. The therapeutic Ad also encodes mutated human IL-4 that binds to the physiological IL4R/IL13R without interacting with IL13Rα2, thus inhibiting potential binding of mhIL-13-PE to normal brain cells. Using intracranial GBM xenografts and syngeneic mouse models, we tested the Ad.mhIL-4.TRE.mhIL-13-PE and two protein formulations, hIL-13-PE used in clinical trials (Cintredekin Besudotox) and a second-generation mhIL-13-PE. Cintredekin Besudotox doubled median survival without eliciting long-term survival and caused severe neurotoxicity; mhIL-13-PE led to ∼40% long-term survival, eliciting severe neurological toxicity at the high dose tested. In contrast, Ad-mediated delivery of mhIL-13-PE led to tumor regression and long-term survival in over 70% of the animals, without causing apparent neurotoxicity. Although Cintredekin Besudotox was originally developed to target GBM, when tested in a phase III trial it failed to achieve clinical endpoints and revealed neurotoxicity. Limitations of Cintredekin Besudotox include its short half-life, which demanded frequent or continued administration, and binding to IL4R/IL13R, present in normal brain cells. These shortcomings were overcome by our therapeutic Ad, thus representing a significant advance in the development of targeted therapeutics for GBM.