Google DeepMind Announces Secure, Server-Side Memory for Private AI Compute

Google DeepMind has unveiled a new technical capability for its Private AI Compute platform, enabling persistent, cross-device AI memory with on-device privacy standards. This breakthrough addresses a longstanding challenge in modern AI: how to provide continuous assistance across devices while maintaining strict privacy standards typically limited to on-device processing.

With this new capability, a secure digital vault in the cloud will function as a persistent memory layer. The information needed to assist users is stored in dedicated, encrypted storage, while cryptographic keys are held exclusively on personal devices, ensuring data remains inaccessible to anyone else, including Google.

This innovation allows for seamless experiences, such as resuming complex conversations between mobile and web devices or accessing previously viewed content across different devices. To build user trust, Google is publishing a tamper-proof public record of its server software, enabling devices to verify the software's authenticity before sending personal data. Additionally, the company is providing an update on its technical methods, including the results of an independent audit by a leading cybersecurity firm.

Google DeepMind invites the broader privacy community to review the updated Private AI Compute Technical Brief, system architecture, security proofs, and verification protocols. This research was co-developed by Google DeepMind, Platforms & Devices, Core, and Cloud teams, with executive sponsorship from Four Flynn, Jay Yagnik, and David Kleidermacher.

Source: deepmind