Human Misperception of Generative-AI Alignment: A Laboratory Experiment