Updates to Consumer Terms and Privacy Policy

Can users poison the future training dataset by ending every chat with a dissatisfied feedback even though the chat helped them? Or be even more malicious and steer the chat into destructive behavior and then give very positive feedback?