Background: Adaptive immune resistance (AIR) is pivotal to triple-negative breast cancer (TNBC) progression. Although immune checkpoint inhibitors (ICIs) shed new light on reversing AIR, the treatment effect and predictive markers of immunotherapy remain controversial. The present study focused on epigenetic regulation for TNBC AIR and explored whether this could be clinically targeted. Methods: Through screening an epigenetic gene set in TNBC patients from The Cancer Genome Atlas (TCGA), ARID1A was pinpointed as most significantly correlated with AIR. Humanized immune system (HIS) mice were adopted for TNBC transplantation, identifying CD8⁺ T cells as most significantly correlated with TNBC ARID1A. To probe the mechanism underlying TNBC ARID1A deficiency-induced CD8⁺ T cell malfunction and AIR, RNA-seq of both TNBC patients and TNBC cell lines were analyzed and the candidate mediator gene, PD-L1, was identified and verified through western blot, qPCR, IHC, dual luciferase reporter assay and cell co-culture. Chromatin immunoprecipitation and ATAC-seq further demonstrated direct binding of ARID1A to NPM1 promoter to regulate PD-L1 expression. Finally, HIS mice were treated with ICI and the clinical trial CTR20191353 was retrospectively analyzed. Results: Among a series of epigenetic modulators, we first observed that ARID1A, a core subunit of the chromatin remodeling complex SWI/SNF, demonstrated highest correlation with AIR in TNBC patients. This was verified in HIS TNBC mice and a patient cohort from Fudan University Shanghai Cancer Center by higher malignancy observed in ARID1A-deficient TNBC. Both bulk and single-cell RNA-seq in TNBC patients uncovered an immunosuppressive microenvironment induced by TNBC ARID1A deficiency, and CD8⁺ T cells was identified as the immune cell type most significantly correlated with TNBC ARID1A. CD8⁺ T cell malfunction and AIR in ARID1A-deficient TNBC was induced by upregulated PD-L1, but ARID1A did not directly regulate PD-L1 expression. We found that ARID1A bound to the NPM1 promoter and that ARID1A deficiency increased NPM1 chromatin accessibility as well as gene expression, further activating PD-L1 transcription. In HIS mice, ICI demonstrated the potential to reverse TNBC ARID1A deficiency-induced AIR manifested by reducing tumor malignancy and activating antitumor immunity. In CTR20191353, a phase 1b clinical trial of pucotenlimab combined with chemotherapy for TNBC, patients with ARID1A deficiency demonstrated significantly better prognosis and the longest progression-free survival was found in ARID1A-low/PD-L1-high group. Conclusions: In AIR epigenetics, TNBC ARID1A deficiency induced AIR via the ARID1A/NPM1/PD-L1 axis, leading to poor outcome. While ARID1A deficiency provided survival advantage for TNBC cells, it simultaneously left the Achilles' heel which could be targeted with ICI.