Looking into how AI voice synthesis works and there are models that don't rely on tons of training data and can be done at home
And completely avoids Data centers and whole sale data theft. It just requires people willing to train it.
All the ethical problems of it are avoided with proper use.