The old two word captchas were one word that Google knows in order to test if you're human, and one word scanned from Google's book scanning program that their algorithms failed to properly OCR, meaning for the second word you could type in whatever you wanted and you would pass the captcha
Sites were allowed to use recaptcha for free because their users were actually doing work training neural nets to read books better, if a large percentage of their users are saying every unknown scan is the n-word, I could see why Google wouldn't want them having access to it