Gliomas are the most common cancerous brain tumors, constituting the primary cause of death associated with brain tumors. Their widespread occurrence and the complexities involved in their treatment highlight the urgent need for continued research and innovation to improve diagnosis, treatment, and patient survival. Manual segmentation of gliomas in MRI (magnetic resonance imaging) scans is a labourous job and subject to variability, necessitating more efficient and accurate automated methods. This paper presents a pioneering approach by integrating a dynamic feature-gated attention mechanism into a 3D U-Net architecture for the semantic segmentation of gliomas in volumetric MRI scans. Our model combines the robust U-Net architecture with adaptive, selective focusing capabilities of feature-gated attention, effectively capturing both local and global dependencies within the volumetric data. The dynamic feature-gated attention mechanism updates attention weights during training, allowing the model to adjust feature importance based on the input data. Integration of this mechanism enhances feature discernment and segmentation accuracy. The experimentation evaluation conducted on the BraTS 2019 dataset, including measures like sensitivity, specificity, precision, IoU score, accuracy, and Dice Similarity Coefficient (DSC), demonstrates promising results. Our method achieves dice scores of 0.90 for the whole tumor (WT), 0.84 for the core tumor (CT), and 0.84 for the enhancing tumor (ET), along with an impressive accuracy rate of 98%.