I tried the example prompt and I got a copypasta filled with racial slurs. Is this like the tay.ai situation where user input is being added to training data?