Artificial intelligence (AI) is rapidly reshaping how law enforcement agencies approach crime prevention and public safety. One of the most prominent and contested applications is predictive policing, which leverages machine learning algorithms to analyze historical and real-time data in order to forecast future criminal activity. While these systems offer the promise of improved resource allocation and proactive intervention, they also raise significant ethical, legal, and social concerns. This paper critically examines the development and deployment of AI-driven predictive policing by synthesizing academic literature and presenting a scenario-based analysis of system implementation in a fictional urban setting. Key technologies discussed include risk-scoring algorithms, geospatial crime mapping, and behavioral prediction models. Recurring issues such as algorithmic bias, data opacity, and the disproportionate targeting of marginalized communities are explored in depth. The Metroville case scenario illustrates how these concerns manifest in practice highlighting risks related to over-policing, loss of public trust, and governance gaps when AI tools are deployed without appropriate oversight. Building on this analysis, the paper proposes a set of guiding principles for the ethical integration of AI into policing: transparency, fairness, community involvement, and independent auditing. The findings emphasize that technical efficacy alone is insufficient; legitimacy, accountability, and inclusivity must be central to any AI adoption strategy in public safety contexts. This work contributes to the growing discourse on responsible AI by offering practical recommendations for policymakers, technologists, and law enforcement agencies seeking to align innovation with democratic values and human rights.