Dam breach modelling is a critical component of dam safety programs, supporting emergency preparedness, consequence assessment, and regulatory decision-making. Traditional practice relies on deterministic approaches, where a single set of breach parameters represents a “worst-case” failure. However, key parameters—such as breach geometry, formation time, and erosion behaviour—are highly uncertain and cannot be directly observed. To address this, deterministic analyses often combine conservative assumptions, which can lead to exaggerated flood estimates and simplified, binary representations of risk. This paper presents a practical framework for probabilistic dam breach modelling that complements deterministic methods. Using Monte Carlo simulation, uncertain parameters are represented as statistical distributions and sampled across thousands of realizations. The resulting ensemble enables estimation of exceedance probabilities for peak discharge and flood extents. The methodology uses HEC-RAS with automated workflows to support efficient largescale simulation. Key considerations include development of parameter distributions from sensitivity analyses, use of truncated models to improve efficiency, and evaluation of statistical convergence. Results show that deterministic “worst-case” scenarios may overestimate flood consequences relative to probabilistic outcomes. By quantifying likelihood, probabilistic modelling provides a more transparent, risk-informed basis for decision-making. The approach supports improved risk communication, emergency planning, and infrastructure investment, and aligns with the Canadian dam safety community’s transition toward risk-informed frameworks.