Train on what is on your screen

It does not. What is on screen gets used for the thing you asked for and goes nowhere else afterwards. None of it is kept, and none of it goes into making a model better. That answer stays the same whichever mode you have switched on.

What happens

The window is read for the task in front of you. When the task ends, nothing about what was on screen is kept.

Worth knowing

If you would rather your requests never touched our account, you can bring your own provider key and pay your provider directly.

True in every mode.