Google's Private AI Compute Team has announced a technical update to its Private AI Compute architecture that will enable persistent, cross-device AI memory while upholding the strict privacy standards typically associated with on-device processing. In a blog post published on the Google DeepMind blog, the team said the update resolves a longstanding dilemma in modern AI: how to give an assistant long-term continuity across devices while keeping data as protected as local processing.

According to the company, AI is becoming more capable and intuitive — remembering what matters, understanding the world around the user, and acting at their direction. Google said privacy and trust are core to making that possible, ensuring data stays private and protected as AI systems evolve to provide more continuous assistance across devices.

Bringing on-device privacy to cloud-scale memory

The new capability centers on a persistent memory layer that will function like a secure digital vault in the cloud. Under this model, the information needed to assist a user is sealed within dedicated, encrypted storage, while the cryptographic keys required to unlock it are held exclusively on the user's personal devices — ensuring the data is inaccessible to anyone else, even Google.

When an AI model needs to access information to assist a user, an authenticated, end-to-end encrypted channel connects the device to a protected, isolated environment in the cloud. Google refers to that space as a "secure enclave." It temporarily decrypts the user's data in isolated memory to handle the request, saves any new context, and immediately encrypts it again, keeping the information private "as if it never left your device."

The architecture combines hardware-enforced secure enclaves, encrypted channels, and per-user databases shielded by device-derived encryption keys. Google said this combination keeps data fully private and under the user's control.

The company framed the change as necessary to meet the computing needs of the AI era. Local, on-device processing has historically been the gold standard for privacy, but frontier AI models often require far more computing power than any single device can provide. Bringing advanced AI to personal assistants, Google said, means solving how to tap into the power of the cloud while ensuring personal data remains as protected as if it never left the device.

To that end, Google previously introduced its Private AI Compute platform, which allows users to process complex tasks in hardware-isolated cloud enclaves. Until now, that technology — along with similar solutions across the industry — was strictly "stateless," meaning it wiped all context the moment a task ended. Google argued that workarounds, such as having an AI save a list of personal facts and preferences, are not enough to support the rich, continuous experiences people expect from personal AI. Enabling that level of assistance required engineering a way for cloud-scale AI to securely retain context over time and across devices.

That stateless constraint was never unique to Google: hardware-isolated cloud processing has been adopted across the industry for running sensitive workloads with cloud-scale resources, and as long as such systems wiped context after each task, cloud-based AI offered no durable equivalent of the memory local software provides. The move to persistent server-side memory is therefore a notable architectural shift for cloud AI — and one Google argues is a prerequisite for personal assistants that maintain genuine continuity across devices.

Building trust, looking ahead

Google illustrated the potential use cases: pulling up assembly instructions on a laptop that were previously viewed through smart glasses, or resuming complex conversations between mobile and web. Private AI Compute is designed to make that kind of seamless assistance possible while keeping the pieces it needs to remember safely locked away. The company acknowledged, however, that user trust in the system's privacy is equally important.

Building that trust starts with transparency, according to Google. Alongside its updated technical whitepaper, the company is publishing a tamper-proof public record of its server software. Devices running Private AI Compute will be able to verify that the software is authentic and unaltered before sending any personal data.

Google is also providing an update on its technical methods, including the results of an independent audit by a leading cybersecurity firm, which it did not name. By sharing these resources, the company said it invites the broader privacy community to verify Private AI Compute's protections.

"Adding private, persistent memory to Private AI Compute shows how deeply personal assistance can be private by design," the company wrote, inviting the community to review the updated Private AI Compute Technical Brief as well as its system architecture, security proofs, and verification protocols.

Several practical details remain open: Google did not name the auditing firm, and the announcement did not specify when the persistent memory capability will reach users or which devices will support it first. Those gaps put added weight on the materials Google has made public, which the company says are sufficient for outside privacy researchers to verify its protections directly rather than rely on Google's own assurances.

Acknowledgements

The research was co-developed by Google DeepMind, Platforms & Devices, Core and Cloud teams. Google also thanked Four Flynn, Jay Yagnik, and David Kleidermacher for their executive sponsorship of the work.

The announcement follows Google DeepMind's related post on piloting the world's first double-blind AI evaluations.