Google DeepMind announced persistent server-side memory for Private AI Compute, published tamper-proof software record, and released updated technical whitepaper and independent audit results.
Following·Day 2·first covered 23 Sep 2026·6 sources
Google DeepMind Adds Persistent Encrypted Memory to Private AI Compute
The gist
The update allows persistent AI context across devices without giving the cloud provider access to user data, a combination the prior stateless design could not support. Devices can verify the software is authentic and unaltered against the public record before sending personal data, enabling external accountability.
The full picture
Google DeepMind has added a persistent memory layer to its Private AI Compute platform, storing user data server-side in encrypted cloud storage while keeping the cryptographic keys exclusively on users' personal devices, making the data inaccessible to Google. The memory runs inside isolated hardware-enforced secure enclaves that temporarily decrypt data to handle a request and immediately re-encrypt it. The system enables cross-device continuity, such as resuming a conversation started on smart glasses when switching to a laptop. Previously, Private AI Compute and similar industry solutions were strictly stateless, wiping all context after each task. To support external verification, Google is publishing a tamper-proof public record of its server software and an updated technical whitepaper, and has shared results from an independent cybersecurity audit covering 2025 and 2026.
How it developed
Sources
The Neuron
helpnetsecurity.com
1 more source
The daily email
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free