“The official ChatGPT prompting guide recommends setting boundaries for the model to avoid creating extra work or unintended actions.”