Another thing to potentially look into is the headers being sent:
As for text, I guess your best bet is to either limit exposure of it or intentionally poison the data. It's a little bit harder to do this in writing I guess, but you could still try by creating unusable fictitious accounts of characters named "Biden" or "Boris" doing increasingly ridiculous things. Any politically-stark moderator would probably remove your data before it hits the model, and if it does there's a good chance it will be flagged as problematic.