Volcanic thermal anomalies are commonly monitored using advanced optical satellite sensors, enhancing the detection of renewed volcanic activity. Traditionally, fixed-threshold hotspot detection algorithms have been widely applied to identify these anomalies, effectively minimizing false alarms. However, the mapping of lava flows and monitoring of volcanic activity, which is essential for hazard mitigation and understanding the behavior of active volcanoes, has been further improved through the use of Machine Learning techniques. These methods allow for the rapid processing of large datasets, making them especially valuable for volcanic studies. Here, a Machine Learning approach based on a Random Forest algorithm, designed and implemented on Google Earth Engine, using data from the Sentinel-2 multispectral sensor (S2-MSI), is applied to detect and accurately map lava flows from the 2023-2024 eruption in Sundhnúkur, Iceland. Despite gaps in satellite coverage due to technical issues or adverse weather, the flow maps generated by the algorithm closely align with the actual lava flow fields. The results demonstrate that the Random Forest model, despite not being trained on this study area, exhibits strong generalization capabilities and high sensitivity to subtle volcanic thermal anomalies.