CAR T cells have shown clinical efficacy for acute lymphoblastic leukemia, but this therapy has not been effective for acute myeloid leukemia (AML), and other treatment options are needed. Theoretically, CAR-NK cells have a more favorable toxicity profile compared to CAR T cells, especially in avoiding adverse effects such as cytokine release syndrome. However, the clinical evidence for this has not yet been reported. In the current study, we tested the safety of CD33-CAR NK cells in patients with relapsed and refractory AML. At doses up to 5 × 10 (5 billion) cells per patient, no significant adverse effects were observed. CAR NK-92 cells can be produced at much lower cost compared to CAR T cells, and we believe after being optimized, they will be widely accessible for the treatment of cancer.