The occurrence of severe adverse events (SAEs) in patients with Duchenne muscular dystrophy (DMD), X-linked myotubular myopathy (XLMTM), and other neuromuscular diseases treated with adeno-associated virus (AAV) constructs has prompted studies to improve the safety and efficacy of gene therapy. Physicians have weighed the medical tenet of “first, do no harm” against the perspective of patients with progressive life-threatening conditions who may accept greater risk. Regarding SAE pathogenesis, discussion has focused on total AAV exposure and patient mutations more likely to induce immunity, while stressing the limitations of animal models in predicting adverse events. Therapeutic strategies for reducing side effects have employed more myotropic AAV serotypes and efficient transgenes. Other recommendations include excluding certain DMD gene mutations associated with SAEs and substituting less immunogenic transgenes such as utrophin (DMD) and myotubularin-related protein (XLMTM). For the sake of preclinical studies, emphasis has been placed on outbred rodents and larger animals that better predict immunity. Here, the absence of side effects in canine DMD and XLMTM models might be explained partly by phenotypic differences between affected humans and dogs. Specifically, dystrophin- and myotubularin-deficient dogs exhibit milder lesions, including less muscle fat deposition and the absence of hepatopathy, respectively, which could lead to reduced immune responses to AAV constructs. To better predict future problems, thought should be given to tracking early subclinical markers of the innate immune response, especially complement activation. Regardless of steps taken to improve the predictive value of animal models for SAEs, some questions will only be answered through human clinical trials after carefully considering the risk-benefit ratio.