They should be marked by default as non-trainable and you should have to opt each individual chat in as you see fit.
The cognitive load to remember to opt out every new chat should not rest on the user.
They should be marked by default as non-trainable and you should have to opt each individual chat in as you see fit.
The cognitive load to remember to opt out every new chat should not rest on the user.
What will any provider get out of running free inference then?
Compliance with the law and therefore the ability to operate at all