Google Introduces Private AI Compute Architecture for Server-Side Memory

Private AI Compute Architecture Update

Google has updated its Private AI Compute architecture to include persistent, server-side memory. This development aims to provide AI systems with long-term continuity across multiple devices.

The new architecture is designed to maintain on-device privacy standards, ensuring user data remains protected even when stored and processed in the cloud.

Secure Server-Side Memory Implementation

The updated system functions by storing user information in a secure digital vault in the cloud. This data is encrypted, and the cryptographic keys required to decrypt it are held exclusively on the user's personal devices.

This design ensures that Google cannot access the user's data, maintaining privacy similar to on-device processing.

Data Processing in Secure Enclaves

When an AI model requires access to user information, an authenticated, end-to-end encrypted channel connects the user's device to a protected, isolated environment in the cloud, referred to as a 'secure enclave'.

Within this secure enclave, data is temporarily decrypted in isolated memory to fulfill the request, new context is saved, and the data is immediately re-encrypted. This process is intended to keep user information private as if it were never processed off the device.