New Google feature lets AI assist you without exposing data
What's the story
Google has launched a new platform, Private AI Compute, to bring the full power of its Gemini AI models to users while keeping their personal data completely private. The system is similar to Apple's Private Cloud Compute and marks a wider industry trend toward privacy-preserving artificial intelligence (AI) that combines cloud intelligence with on-device security.
Technological innovation
What is Private AI compute?
Private AI Compute is Google's latest effort in what it calls "privacy-enhancing AI technology." The platform combines the computational prowess of Gemini cloud models with on-device privacy protections. In simple terms, it allows your device to leverage Google's most powerful AI capabilities—like summarizing recordings or generating contextual suggestions—while ensuring that no one, not even Google engineers, can access your private data.
Security measures
How it works
Jay Yagnik, Google's VP of AI Innovation and Research, explained that Private AI Compute operates within a hardware-secured, sealed cloud environment built on the company's custom Tensor Processing Units (TPUs). The system uses encryption and remote attestation to securely connect your device with this space for isolated data processing. This way, user data remains private even from Google itself.
Performance comparison
Bridging the gap
While on-device AI, like that on Pixel phones, focuses on privacy, it has limitations in processing power. Private AI Compute bridges this gap by letting sensitive tasks leverage cloud-based Gemini models without compromising user privacy. This means smarter and faster responses in tools like Magic Cue and Recorder—now capable of real-time summarization across more languages.
Industry impact
A new era in AI
Google's Private AI Compute marks a major shift toward privacy-first cloud AI, echoing Apple's strategy but powered by Gemini. As AI assistants become more capable and personal, Google's new system ensures that intelligence doesn't come at the cost of user trust. This development is likely to have a significant impact on the future of artificial intelligence technology and its relationship with user privacy.