Google DeepMind has published technical details for persistent server-side memory in its Private AI Compute architecture.
The company says the design is intended to let AI assistants retain context across devices while protecting data through hardware-enforced secure enclaves, encrypted channels and device-derived keys.
Google says the decryption keys remain under the user’s device control and memory is stored in per-user encrypted databases. These security properties are company claims that will continue to be subject to external scrutiny.