AI redefines video analytics for public safety, enabling fast threat detection and response. Traditional systems that rely on human oversight struggle with timely and accurate threat ID. AI-driven video analytics can detect objects, track movements, recognize unusual activity precisely, and provide advanced threat assessment and situational awareness in public spaces. This paper will explore the core capabilities of AI in video analytics: object recognition, anomaly detection, and multi-camera integration, which provide a complete view of high-risk areas. Recent advancements like edge computing enable real-time processing, making AI-driven insights available to security and emergency teams with little delay. These capabilities are particularly relevant for public safety applications like crowd management and incident response at significant events. Ethical considerations like data privacy, algorithmic fairness, and cybersecurity will also be discussed, as well as real-world examples of AI video analytics in major US cities. The paper will conclude with future directions: autonomous aerial monitoring and predictive analytics for proactive threat detection. This paper will argue for ethical, research-driven placement of video analytics to facilitate safer, more sustainable public spaces in the US.