With the rapid growth of digital communication platforms, voice calls and online chat systems have become common mediums for interaction, but they are also increasingly exploited for coordinating unlawful activities. Traditional surveillance approaches based on simple keyword detection often fail to understand context, emotions, and evolving language patterns, leading to high false positives and missed threats. This project proposes an AI-based surveillance system for detecting criminal intent in both voice and chat communication through a unified platform. The system converts voice inputs into text using speechto-text technology and applies natural language processing (NLP) techniques to analyze messages for suspicious intent in real time. If any unwanted or potentially harmful message is detected, it is automatically reported to an admin dashboard for further action. Additionally, the system incorporates a continuous learning mechanism where newly identified keywords or sentence patterns are treated as new data, enabling the model to adapt and improve over time. The proposed system aims to deliver contextaware, real-time threat detection, reduce false alerts, and enhance proactive monitoring in modern communication environments.