New Releases

Apple iOS 27 Leak: Local AI Trained via Gemini, Siri to Route Complex Tasks to Google Cloud

Reading Guide

2 min read

Apple iOS 27 Leak: Local AI Trained via Gemini, Siri to Route Complex Tasks to Google Cloud

According to a detailed report from *The Information*, Apple is deepening its artificial intelligence partnership with Google for the upcoming Apple iOS 27 release. The leak reveals how Apple plans to balance its strict privacy commitments with the intense compute requirements of modern generative AI, focusing heavily on a hybrid on-device and cloud architecture.

Apple iOS 27 Leak: Local AI Trained via Gemini, Siri to Route Complex Tasks to Google Cloud

To power its on-device features, Apple is reportedly using a process called “knowledge distillation” to train its own lightweight local AI model. By utilizing Google’s massive Google Gemini model as a “teacher,” Apple’s engineers are training smaller “student” models that can run locally on iPhones. This approach allows the device to deliver highly responsive, offline AI capabilities without needing to upload sensitive user data to the cloud by default—perfectly aligning with Apple’s long-standing privacy narrative.

To further shrink these models for mobile hardware, Apple has been actively seeking to acquire specialized AI startups. Reports indicate the tech giant recently considered buying Liquid AI, a Cambridge, Massachusetts-based startup renowned for its pioneering work in optimizing neural networks to run efficiently on local, low-power devices.

However, for complex queries that require massive computing power, Apple’s proprietary Private Cloud Compute infrastructure simply won’t have the capacity to handle everything. The report notes that in Apple iOS 27, Siri will automatically route demanding user requests to Google Cloud, utilizing an authorized version of the full-scale Google Gemini model to process the prompt.

To protect user privacy during these cloud transfers, Apple has recently approved the integration of NVIDIA’s “Confidential Computing” technology. This hardware-level security encrypts data and models even while they are being actively processed by the GPU. While confidential computing adds a microscopic delay to AI query response times, it provides a powerful layer of cryptographic protection for sensitive personal information, meaning Apple’s upcoming cloud-AI push will likely rely on a secure combination of Google Cloud and specialized NVIDIA chips.

Previous Huawei Reportedly Testing Multi-Camera Fusion Tech and Custom CMOS to Shake Up Mobile Imaging in 2027 Next vivo S60 Series Launched: 4K Live Photos, Ultra-Thin 7,200mAh Blue Ocean Battery, Starting at 2899 Yuan
C
About cizchu

Senior Technology Editor with 10 years of experience covering mobile technology.

Recommended Articles