Advancing Private AI Compute with secure, server-side memory
deepmind ·
Google DeepMind has announced a technical update to its Private AI Compute platform that introduces secure, server-side memory capabilities. This advancement allows for persistent, cross-device AI continuity while maintaining strict on-device privacy standards. The system functions like a cloud-based digital vault where data is encrypted in dedicated storage, with decryption keys held exclusively on the user's personal devices. When an AI model requires information, it accesses a protected environment via an end-to-end encrypted channel, temporarily decrypting data in isolated memory before re-encrypting it immediately. This architecture resolves the dilemma of providing long-term assistant continuity without compromising the security typically associated with local processing. To build trust, Google is publishing a tamper-proof public record of its server software and sharing results from an independent cybersecurity audit. The update aims to support rich, continuous experiences across multiple devices while ensuring users retain full control over their personal data.
deepmind