Advancing Private AI Compute with secure, server-side memory
Google DeepMind introduces persistent, private server-side memory for its Private AI Compute architecture using secure enclaves and device-held keys.
Google DeepMind details an update to Private AI Compute enabling persistent, cross-device AI memory in the cloud. The system uses a secure enclave and device-derived encryption keys to keep user data encrypted and inaccessible to Google. This allows cloud-scale AI to maintain long-term context while upholding on-device privacy standards.
- Google introduces persistent server-side memory for Private AI Compute.
- Architecture uses hardware-enforced secure enclaves and device-held encryption keys.
- Addresses the cloud scale vs. on-device privacy dilemma for long-term AI assistance.
- Vendors
- GoogleGoogle DeepMind
- Products
- Private AI Compute
- Organizations
- Google DeepMindGooglePlatforms & DevicesCoreCloudFour FlynnJay YagnikDavid Kleidermacher
- AI models
- Private AI Compute
From the original article
3 min read at deepmind.google
Introducing private, server-side memory to Private AI Compute for personal AI.
By Google Private AI Compute Team · deepmind.google
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.