Runtime Application Security Protection (RASP) is crucial in safe-guarding applications against evolving cyber threats. This research presents a novel approach leveraging a fine-tuned BERT (Bidirectional Encoder Representations from Transformers) model as the cornerstone of a robust RASP solution. The fine-tuning process optimizes BERT’s natural language processing capabilities for application security, enabling nuanced threat detection and mitigation at runtime. The developedRASP system harnesses BERT’s contextual understanding to proactively identify and neutralize potential vulnerabilities and attacks within diverse application environments. Through comprehensive evaluation and experimentation, this study demonstrates the efficacy and adaptability of the BERT-based RASP solution in enhancing application security, thereby contributing to the advancement of proactive defense mechanisms against modern cyber threats.