Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study