Google DeepMind details secure server-side memory for Private AI Compute

Google DeepMind announced a technical update to its Private AI Compute platform that introduces a persistent, server-side memory layer designed to provide long-term continuity across devices while preserving on-device privacy standards.

The new architecture adds a secure digital vault in the cloud where user information is sealed within dedicated, encrypted storage. Cryptographic keys required to access the data are held exclusively on the user’s personal devices, ensuring that even Google cannot access the information.

When an AI model needs to retrieve or store context, an authenticated, end-to-end encrypted channel connects the user’s device to a protected, isolated cloud environment called a “secure enclave.” Within this enclave, data is temporarily decrypted to handle the request, new context is saved, and the data is immediately re-encrypted, keeping user information private as if it never left the device.

The system combines hardware-enforced secure enclaves, encrypted channels, and per-user databases shielded by device-derived encryption keys to ensure data remains fully private and under user control.

Until now, Private AI Compute and similar industry solutions operated in a stateless manner, wiping all context after a task ends. This limitation prevented rich, continuous experiences expected from personal AI assistants. The new memory layer resolves this by enabling cloud-scale AI to securely retain context over time and across devices.

To build user trust, Google DeepMind is publishing a tamper-proof public record of its server software and the results of an independent audit by a leading cybersecurity firm. Devices running Private AI Compute will be able to verify the software’s authenticity before sending personal data.