Evaluating safeguarding in multi-turn conversations with AI assistantswith Franziska Sofia Hafner, Luc Rocher1 versionW_h4gruktvv1 · current