Human leukocyte antigen G (HLA-G) is a non-classical MHC class Ib molecule with potent immunomodulatory properties that distinguish it from classical HLA class I antigens. Unlike HLA-A, -B, and -C, HLA-G exhibits limited allelic polymorphism and a restricted tissue distribution under physiological conditions, being primarily expressed at the fetal-maternal interface on extravillous trophoblasts, as well as in immune-privileged sites such as the cornea and thymus. The HLA-G gene generates seven distinct isoforms through alternative splicing, including four membrane-bound (HLA-G1 to -G4) and three soluble forms (HLA-G5 to -G7). Functionally, HLA-G exerts its immunosuppressive effects by binding to inhibitory receptors, primarily ILT2 (LILRB1/CD85j) and ILT4 (LILRB2/CD85d), which are expressed on natural killer (NK) cells, T lymphocytes, dendritic cells, and macrophages. This interaction inhibits cytotoxic T lymphocyte and NK cell-mediated cytolysis, promotes regulatory T cell expansion, and skews immune responses toward an anti-inflammatory Th2 profile. While HLA-G expression is essential for establishing maternal-fetal immune tolerance and preventing fetal rejection during pregnancy, its de novo or upregulated expression in various malignancies, including ovarian, breast, cervical, and colorectal cancers, facilitates tumor immune evasion and correlates with poor prognosis. Notably, emerging evidence positions HLA-G as a novel immune checkpoint molecule, with soluble HLA-G (sHLA-G) and HLA-G-bearing exosomes being investigated as potential liquid biopsy biomarkers. Therapeutic strategies targeting the HLA-G/ILT axis, including monoclonal antibodies such as TTX-080 and BND-22, are currently under clinical evaluation for cancer immunotherapy. The unique dual role of HLA-G in both physiological tolerance and pathological immune evasion underscores its potential as a diagnostic biomarker and therapeutic target.