Google drops VaultGemma, an LLM built to forget your secrets
Google and DeepMind just open-sourced VaultGemma, a 1B-parameter model trained from scratch with differential privacy. By mathematically injecting noise into training, it ensures the model never memorizes sensitive sequences. While privacy constraints pull its benchmark performance back to roughly GPT-2 levels, it provides the first empirical scaling laws for provably private AI.