Background: Despite major therapeutic advances of chimeric antigen receptor (CAR) T cells directed against CD19 in relapsed and refractory non-Hodgkin lymphoma (NHL), two thirds of patients experience relapse within the first two years. Nearly one third of these relapses are associated with CD19 antigen-loss escape thus leveraging alternate CAR targets may improve outcomes. CD47, an anti-phagocytic “don’t eat me” receptor known to be overexpressed in NHL, is an ideal CAR target as it is potentially indispensable for tumor progression. Methods: CD47-CAR binding domains were engineered either as a camelid single variable heavy chain nanobody (CD47-VHH-CAR) or as the CD47 binding domain of cognate receptor SIRPα (SIRPα-CAR), each combined with a CD28 costimulatory domain and a CD3ζ activation domain with two mutated ITAMs (1XX). In vitro co-culture assays were performed with lymphoid mouse tumor cell lines BM185 and A20. For in vivo experiments, syngeneic BALB/c mice were sublethally irradiated for lymphodepletion then injected intravenously with 1x10⁵ luciferase⁺ BM185 cells. Two days later, mice were injected with 2x10⁶ CAR T cells, and bioluminescence imaging was performed to monitor tumor progression. For bio-localization experiments, CAR T cells were co-transduced with a separate vector encoding luciferase. Results: CD47-CAR T cell expansion in vitro was hindered due to fratricide, which can be rescued by CD47 CRISPR knockout. Both CD47-VHH-CAR and SIRPa-CAR T cells exhibited potent and specific in vitro lysis of BM185 and A20 cells. CD47-VHH-CAR T cells showed transient anti-leukemia activity, however, survival was not significantly prolonged. T cell bio-localization analysis revealed that, compared to CD19-CAR T cells, CD47-VHH-CAR T cells exhibited prolonged retention in the thorax with minimal bone marrow trafficking, suggesting CD47 antigen-specific pulmonary sequestration, and inferior expansion. SIRPα-CAR T cells engineered to secrete pro-inflammatory cytokine IL-18 significantly improved CAR T cell expansion, resulting in bone marrow trafficking and clearance of BM185 tumor. However, recipients of IL-18-secreting SIRPa-CAR T cells experienced lethal CAR T cell-related toxicity. To improve the safety of CD47-CAR T cells, we engineered an NFAT-responsive vector expressing the CD47-CAR modified with E. coli-derived dihydrofolate reductase drug destabilization domain to limit CD47-CAR expression to T cells exposed to the stabilizing antibiotic trimethoprim and receiving a licensing NFAT signal from a CD19-CAR. Conclusions: In vivo activity of CD47-targeting CAR T cells is limited by thoracic retention and limited expansion, which can be partially overcome by overexpression of IL-18, though enhanced toxicity precludes survival prolongation. Our findings suggest that strategies to spatiotemporally restrict CAR expression or activation will be necessary to achieve safe and efficacious CD47-CAR T cell activity.