LLMs work by always predicting the next most likely token and LLM detection works by checking how often the next most likely token was chosen. You can tell the LLM to choose less likely tokens more often (turn up the heat parameter) but you will only get gibberish out if you do. So no, there is not.

What do you mean by email aliases?

That’s Japanese not Chinese.

The probability of getting a finite number is pretty much zero.

For any range [0; n], where n is finite, there are always infinitely many numbers larger than n, so the probability of getting a number in said range is n/(n+infinity). I feel very confident in saying that something with that probability will never happen.

They sell ad spots based on their data about you - IMHO that’s very different from outright selling your data.

And