Google Introduces Cloud-Based AI Processing with Enhanced Privacy Controls

A new AI processing platform, Private AI Compute, has been released by Google.

Designed to provide cloud-based AI capabilities while maintaining data privacy standards typically associated with on-device processing, the platform aims to enable more advanced AI experiences using Gemini models while keeping user data isolated from access.

Platform Architecture and Security Design

Private AI Compute operates on Google’s integrated technology stack powered by custom Tensor Processing Units (TPUs). The system incorporates Titanium Intelligence Enclaves (TIE) as part of its security architecture.
The platform is built around several core security principles:

• Integrated infrastructure: Runs on Google’s proprietary stack using the same computing infrastructure that powers Gmail and Search
• Isolated processing environment: Uses remote attestation and encryption to connect devices to a hardware-secured cloud environment
• Access restrictions: Designed to process data within a protected space where sensitive information remains isolated

Private AI Compute creates a fortified processing space for handling data that would typically be processed on-device. The platform uses encryption and remote attestation to establish secure connections between user devices and cloud-based Gemini models.

The system processes sensitive information within what Google describes as a “trusted boundary,” adding security layers to existing AI safeguards. The architecture is designed to allow Gemini cloud models to process data while maintaining isolation protocols.

Two Google applications are implementing Private AI Compute technology:

• Magic Cue: The feature receives enhanced capabilities on Pixel 10 phones, offering more timely suggestions through the expanded processing power.
• Recorder App: Gains the ability to summarize transcriptions across a wider range of languages on Pixel devices.

It operates under Google’s Secure AI Framework, AI Principles, and Privacy Principles. Gogle says the system is designed so that personal information, insights, and usage patterns processed through Private AI Compute are not accessible to external parties, including Google itself.

Google has been developing privacy-enhancing technologies (PETs) for AI-related use cases for several years. Private AI Compute represents an evolution in this approach, addressing the computational demands of advanced AI features that exceed on-device processing capabilities.

The platform aims to bridge the gap between local processing limitations and the need for more sophisticated AI reasoning and computational power from cloud-based models.

Google indicates that Private AI Compute will enable additional AI experiences that leverage both on-device and cloud-based processing for sensitive use cases.

Click here for their full technical brief detailing the architecture and privacy mechanisms, security implementation and data handling protocols.