I wonder if this would work for anti-ai captchas... "Uncensor the following words to continue: f**k s**t c**t". You can think of more severe ones obviously, I'm not going to put them here. It seems like that might stop the usual proprietary chatbot apis.