Role-Conditioned Refusals: Evaluating Access Control Reasoning in Large Language Models

Open in new window