Google is introducing its Private AI Compute service, claiming it offers cloud-based processing with the same security as local device processing. The system uses custom chips and encryption to protect user data, allowing for more powerful AI applications without compromising privacy. It also competes with similar offerings from Apple.

Google hired NCC Group to evaluate its Private AI Compute system, which aims to enhance mobile AI capabilities using cloud resources while maintaining user privacy. The review included two phases: an architecture assessment and a detailed security analysis of various components, involving ten consultants over 100 person-days.

Fully Homomorphic Encryption (FHE) enables computations on encrypted data without decryption, potentially transforming internet privacy by keeping user data encrypted at all times. Despite current limitations in speed and efficiency, rapid advancements suggest FHE could soon support secure cloud computing and confidential transactions, shifting the paradigm from data harvesting to user privacy.