Industrial cities concentrate water, gas, petrochemical, utility, and process pipelines within operating environments where a small undetected leak can escalate into resource loss, soil or groundwater contamination, atmospheric emissions, fire risk, and prolonged service disruption. This review critically examines smart leakage detection systems as an environmental-safety capability for Saudi Arabia's industrial cities under Vision 2030. The literature is synthesized across water distribution, natural-gas and hydrocarbon pipelines, Internet of Things sensing, machine learning, and digital-twin approaches. Recent studies show a clear shift from single-sensor threshold alarms toward multivariate, data-driven and hybrid systems that combine acoustic, vibration, pressure, flow, supervisory control data, geospatial context, and physics-informed models. Yet high reported classification accuracy does not automatically translate into environmental protection because field performance is conditioned by leak rarity, weak signals, transient operations, sensor drift, corrupted telemetry, localization uncertainty, and false-alarm burden. Saudi evidence further indicates that industrial environmental monitoring must account for local meteorology, aridity, heat, dust, connectivity, and governance conditions. The review therefore proposes an integrated architecture in which robust sensing, secure edge analytics, anomaly detection, localization, digital twins, operator verification, and automated containment form a closed safety loop. It concludes that the strongest pathway for Saudi industrial cities is not a single algorithm but a validated, multisensor, risk-based system whose success is measured by detection latency, released volume avoided, localization error, alarm credibility, response time, and auditable environmental outcomes.