well we have all now seen how these models solve such problems, by stealing and training on their users' draft notes on them, without their consent