Adversarial Attack Makes ChatGPT Produce Objectionable Content

There is no clear way to beat the attacks and other Large Language Models are vulnerable too, say computer scientists.