The rise of 5G networks has increased cyberattack complexity due to its decentralized, software-driven architecture. Existing machine learning (ML) solutions often lack real-time user guidance and interpretability. This work presents a hybrid assistant integrating multiple ML models with a Conversational AI interface for real-time prediction of over seven distinct attack types, achieving over 85% combined accuracy. The system delivers conversational explanations, and actionable guidance to facilitate rapid threat identification and informed responses. Experimental results demonstrate enhanced predictive accuracy, interpretability, and user engagement. Key contributions include the development of an integrated multi-model ML and conversational AI framework for interpretable cyberattack mitigation and comprehensive validation of its real-time effectiveness.