SafetyPrompts: a Systematic Review of Open Datasets for Evaluating and Improving Large Language Model Safety