Wine quality assessment has traditionally relied on sensory evaluations by expert wine tasters. This process is inherently subjective and prone to inconsistencies. Poor-quality wines pose serious health risks, including digestive and liver complications. These are often caused by high volatile acidity and chemical additives. Despite these health concerns, there is no globally standardized system for evaluating wine quality. Machine learning presents a promising solution. It enables automation and standardization of wine quality assessment. In this study, a Tuned Random Forest Classifier was utilized to predict wine quality, with a focus on identifying poor quality wines using the UCI Wine Quality dataset. Feature engineering, minority class balancing through oversampling using resampling with replacement, and hyperparameter tuning to mitigate overfitting were used. The proposed model achieved a 97.70% test accuracy, 97.9% cross-validation accuracy, and a 0.977 F1-score in identifying poor quality wines. Volatile acidity was identified as the most important feature for poor quality detection. These results highlight the potential of machine learning to enable global alcohol safety and quality control.