In a real-time deep learning model can be built to detect cyberbullying comments within social media. Such a model is expected to process user comments in real time, tag the possible harmful content in order to give instant feedback to the user. This mechanism of instantaneous response pushes users to introspect about the language used and create awareness and responsibility within their interaction online. It introduced an adaptive learning framework that enhances the effectiveness of the system and users could report false positives and false negatives. User-generated feedback was important in fine-tuning the accuracy of the model so that it could better understand the nuances of language and context, which may vary widely from one community to another and from one platform to another. It will further be open to constant trends of changing language use patterns and new trends on cyberbullying. An overarching goal of this project would be to ensure that an in-time deep-learning-based scheme is aimed toward detecting cyberbullying cases happening around the comment presented with the social media blog. User engagement data will be integrated into continuous model retraining so that the system is sensitive not only to changes in the trend of cyberbullying but also in trends of language use. Dynamism in the feedback loop reduces misclassification and enhances general usability experience and safety of the online world.