<p>The 2023 Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) update refines the atypia of undetermined significance (AUS) category by subclassifying it into AUS-nuclear atypia (AUS-N) and AUS-other atypia (AUS-O) based on risk of malignancy, to enhance clarity of diagnosis and patient management. This study evaluates the impact of AUS subclassification on malignancy risk prediction in thyroid nodules. A total of 118 AUS cases were analyzed for subclassification using the nuclear scoring (NS) system to evaluate nuclear features, along with the ultrasonography-based Korean Thyroid Imaging Reporting and Data System (K-TIRADS) for risk stratification, and <i>BRAF</i> V600E mutation testing. Logistic regression and ROC curve analyses were used to identify predictors of malignancy and to assess model performance. The AUS category was divided into AUS-N and AUS-O, with AUS-N having a significantly higher risk of malignancy (ROM) (78.1%) compared to AUS-O (27.3%). The subcategories AUS-N1, N2, and N5 showed significantly high ROM (96.9%, 71.7%, and 90.0%), whereas AUS-N3 and N4 showed lower ROM (20.0% and 33.3%). The NS system standardized the assessment of nuclear atypia, reducing interobserver variability and improving diagnostic reproducibility. <i>BRAF</i> V600E mutation, present in 28.1% of AUS-N cases but absent in AUS-O cases, was a strong predictor of malignancy. Models integrating imaging, detailed cytologic subclassification, and molecular findings achieved high specificity (81.0–86.5%) but moderate sensitivity (58.0–61.3%). These findings support the use of AUS subclassification and molecular testing for <i>BRAF</i> V600E mutation to improve ROM prediction and are consistent with the 2023 TBSRTC emphasis on tailored risk stratification.</p>