The administration of allogeneic natural killer (NK) cells following a lymphodepleting chemotherapy regimen is emerging as a well-tolerated therapeutic approach in the management of various malignancies. Contrary to the expected complications of allogeneic T cell therapy, there remains no evidence of graft-versus-host disease (GVHD) mediated by NK cells in numerous clinical trials. On the contrary, preclinical and clinical studies suggest that NK cells do not induce GVHD and in fact may prevent its development following allogeneic hematopoietic cell transplantation (HCT). In this study, we sought to determine the maximum tolerated dose of non-HLA-matched donor NK cells derived from peripheral blood and ex vivo expanded using a novel feeder cell platform. In a single-center Phase I clinical trial using a 3×3 design, 9 subjects each received 2 infusions of NK cells 2 weeks apart following a preparative regimen of cyclophosphamide (60 mg/kg i.v.) and fludarabine (25 mg/m2/day i.v for 5 days). No exogenous cytokines were administered. NK cells were administered at 3 dose levels: 1×107/kg, 2.5×107/kg, and 5×107/kg. Three subjects had myelodysplastic syndrome (MDS) or acute myelogenous leukemia (AML), and the other 6 subjects had colorectal carcinoma. Recipients were monitored over a 4-week period for GVHD as well as other adverse events and for persistence of donor NK cells in systemic circulation. Disease assessment was started at 28 days following the first NK cell infusion and continued until postinfusion day 100 or disease progression. In all 9 study subjects, there was no occurrence of GVHD and no dose-limiting toxicities that would warrant cohort expansion at any of the 3 planned cell dose levels. Low-level donor NK cell persistence was observed up to 4 weeks after the first NK cell infusion at all dose levels. The best observed response was a complete response with incomplete platelet recovery in a MDS subject who experienced disease relapse after prior allogeneic HCT. Other responses were stable disease in 1 subject with MDS and 2 subjects with colorectal cancer up to postinfusion day 100. This off-the-shelf, third-party NK cell product can be administered safely without inducing GVHD and exhibits in vivo persistence promoted by preparative lymphodepletion alone. The observed clinical responses could be enhanced by administration of exogenous cytokine support, as well as complementary approaches that promote NK cell function in the tumor microenvironment.