Encrypted Inference Leaves the Lab
Google released an open-source compiler intended to make homomorphic encryption practical for private AI inference. The concrete change is that a service can process protected data without receiving it in plaintext, potentially opening cloud AI workflows for healthcare, finance and proprietary enterprise data that cannot currently leave controlled environments. This remains one provider implementation; independent benchmarks, usable latency on production workloads and deployments outside Google would confirm or disprove the category.