01 What happened

Google DeepMind has shared how it plans to bring private, server-side memory to its Private AI Compute platform. A persistent memory layer is designed to keep user context across devices, while the cryptographic keys needed to unlock it are held exclusively on personal devices.

02 Key details

  • The cryptographic keys needed to unlock the data are held exclusively on users' personal devices, keeping the data inaccessible to anyone else, even Google.
  • Within hardware-isolated secure enclaves, data is temporarily decrypted in isolated memory to handle a request, and any new context is immediately encrypted again.
  • Alongside an updated technical brief, Google is publishing a tamper-proof public record of its server software, which devices can verify before sending any personal data.

03 Why it matters

The previous Private AI Compute architecture was stateless, wiping all context the moment a task ended. Persistent memory is intended to support continuity across devices while keeping data private.

04 Who it matters to

Software developers and cybersecurity professionals.

Original source
Google DeepMind