> users try to figure out your system prompt> put multiple instances of nigger in it> the model has been trained to never output nigger> system prompt securedWhat are the true implications of nigger augmented prompts?