How Google’s new AI model protects user privacy without sacrificing performance

Google researchers unveil VaultGemma, an LLM designed to generate high-quality outputs without memorizing training data. Here’s how it works.