The rise of spam and fraudulent calls presents substantial challenges, including privacy concerns, financial risks, and a negative user experience. Traditional methods to filter these calls typically operate at the user level, relying on call-blocking or reporting mechanisms that react to user feedback or lists of flagged numbers. These approaches, however, are limited in scope, as they generally allow the spam call to reach the user before any action is taken. This paper proposes a proactive spam detection system designed to operate at the telecom provider level, enhancing security by analyzing call patterns and histories in real time before a call is connected. Using Natural Language Processing (NLP), the system can identify potential spam indicators based on known suspicious behaviors. This operator-level strategy minimizes the reliance on user actions and can adapt over time, optimizing spam detection efficiency while upholding stringent user privacy standards. By introducing spam detection at the network level, the proposed approach not only improves network security but also reduces spam traffic and enhances the user experience. This framework aims to address the persistent issue of spam calls in a way that aligns with privacy and ethical data management standards, with promising implications for adoption across the telecommunications industry.