Google says new cloud-based “Private AI Compute” is just as secure as local processing

In its ongoing quest to integrate generative AI across its product lineup, Google has introduced a new feature named Private AI Compute, aimed at balancing advanced AI capabilities with user privacy. This innovative cloud-based service is designed to improve AI interactions while ensuring that user data remains protected. Google's Private AI Compute operates on what the company describes as a "seamless Google stack," utilizing custom-built Tensor Processing Units (TPUs) that incorporate secure elements. This setup enables devices to establish secure connections to the protected environment via encrypted links, reinforcing data security. The TPUs are enhanced with an AMD-based Trusted Execution Environment (TEE), which is responsible for encrypting and isolating memory from the host system. This design theoretically ensures that no unauthorized parties, including Google, can access user data. An independent review conducted by NCC Group has confirmed that Private AI Compute adheres to Google's stringent privacy standards. Google asserts that the security of this cloud service rivals that of processing data locally on your device. However, the cloud-based solution boasts significantly greater processing power than conventional personal devices, allowing users to leverage Google’s most advanced AI models, including the Gemini series. As the tech giant continues to roll out AI features in devices like its Pixel smartphones, it has emphasized the capabilities of its on-device neural processing units (NPUs). The latest Pixel 10 has seen enhancements in its Gemini Nano models, which facilitate the secure processing of AI tasks at the edge, minimizing the need to share data over the Internet. This upgrade, developed in collaboration with researchers from DeepMind, underscores Google's commitment to advancing AI technology while prioritizing user privacy.