AI CMU Researchers Propose a Simple and Effective Attack Method that Causes Aligned Language Models to Generate Objectionable Behaviors at a High Success Rate
