I argue the opposite position, which is that AI training over private data is great for the world. It is one way for the models to progress, and we receive the benefits of those models a thousand fold.