Google Details Private, Server-Side Memory for AI Assistants

Google says it will add private, server-side memory to Private AI Compute, enabling persistent, cross-device AI memory while retaining privacy standards typically associated with on-device processing. [1] The planned persistent memory layer will place information in dedicated, encrypted cloud storage, while the cryptographic keys required to unlock it remain exclusively on a user’s personal devices. [1]

When a model needs the stored information, an authenticated, end-to-end encrypted channel will connect the user’s device to a protected and isolated cloud environment that Google describes as a secure enclave. [1] Inside that enclave, data will be temporarily decrypted in isolated memory to process the request; new context will then be saved and immediately encrypted again. [1]

Google’s design separates encrypted server-side storage from device-held keys, then confines plaintext handling to an isolated enclave during a request. That architecture attempts to combine the continuity of cloud memory with the privacy posture Google associates with on-device processing, rather than choosing only one model. What the sources don't address: when the capability will reach users, which products will adopt it, or how its security properties will be independently verified.