Underground storage in geologic formations will play a key role in the energy transition by providing low-cost storage of renewable fuels like hydrogen. The sealing qualities of caverns leached in salt and availability of domal salt bodies make them ideal for energy storage. However, unstable boundary shear zones of anomalous friable salt can enhance internal shearing and pose a structural hazard to storage operations. Considering the indistinct nature of internal salt heterogeneities when imaged with conventional techniques like reflection seismic surveys, we develop a method to map shear zones using seismicity patterns in the US Gulf Coast, the region with the world’s largest underground crude oil emergency supply. We developed and finetuned a machine learning algorithm using tectonic and local microearthquakes. The finetuned model was applied to detect microearthquakes in a 12-month nodal seismic dataset from the Sorrento salt dome. Clustered microearthquake locations reveal the three-dimensional geometry of two anomalous salt shear zones and their orientations were determined using probabilistic hypocenter imaging. The seismicity pattern, combined with borehole pressure measurements, and sonar surveys show the spatio-temporal evolution of cavern shapes within the salt dome. We describe how shear zone seismicity contributed to a cavern well failure and gas release incident that occurred during monitoring. Our findings show that caverns placed close to shear zones are more susceptible to structural damage. We propose a non-invasive technique for mapping hazards related to internal salt dome deformation that can be employed in high-noise industrial settings to characterize salt domes used for storage.