The Differential Privacy You're Talking About Is Privacy and Security

Privacy and security are complementary technology. The idea behind differentially private model training is to train the model in a way that guarantees that that can't happen. There's something now called OpenDP, which is an open source toolkit for differential privacy.