Google Unveils HEIR Compiler for 4 Private AI Inference Demos on Encrypted Data

3 articles · Updated · The Keyword | Google Product and Technology News · Aug 14

Google introduced HEIR, an open-source compiler toolchain that converts pre-trained AI models to run inference directly on encrypted inputs without exposing underlying data.
HEIR targets a key barrier to homomorphic encryption: developers typically need specialist cryptographers to adapt software, while Google says it wants the tool to become a one-click path to production use.
Four demo applications compiled with HEIR show current use cases on single-threaded CPUs: recommendation systems, credit-card fraud detection, encrypted network anomaly detection, and hotword recognition.
Since outlining the project in 2023, Google says HEIR has drawn hardware partners including Belfort, Niobium, Cornami and Optalysys, with accelerator-based latency gains planned for later demonstrations.
Google framed HEIR as part of a broader push to make privacy-preserving AI practical in regulated sectors such as healthcare and finance, where data-sharing limits have constrained cloud-based AI services.