Nathan Helm-Burger comments on What should we censor from training data?