Flash floods, which are typically triggered by severe rainfall, are difficult to monitor and predict at the spatial and temporal scales of interest due to large meteorological and hydrologic uncertainties. In France, the AIGA discharge-threshold flood warning system ingests the operational radar-gauge rainfall grids into a simplified distributed hydrologic model at a 1-km² resolution to provide flash flood warnings for small-to-medium (from 10 to 1000 km²) ungauged basins. AIGA produces real-time peak discharge estimates along the river network, which are subsequently compared to regionalized flood frequency estimates of given return periods to characterize the severity of ongoing events. To increase warning lead time while accounting for precipitation forecast uncertainty, AIGA is currently being enhanced to ingest precipitation ensembles from the convection-permitting COSMO-DE-EPS system from the Deutscher Wetterdienst. The 20 precipitation ensemble members of a 21-h horizon at a 2.8-km resolution were ingested in the hydrologic model to produce flow ensembles and probabilistic flash flood warnings for the Meuse and Moselle river basins and for significant events of the 2010-2013 period. The evaluation showed significant improvements in terms of flash flood event detection and effective warning lead-time compared to warnings from the current deterministic AIGA setup, as well as improved forecast skill compared to single-valued persistence benchmarks. Planned enhancements include ingesting other probabilistic products from Météo-France’s convection-permitting AROME model, accounting for hydrologic uncertainties, as well as developing comprehensive observational and post-event damage database to determine decision-relevant thresholds for flood magnitude and probability.