Apple is rebuilding Siri from the ground up for iOS 27, according to a 9to5Mac report. The project moves the assistant beyond simple commands to a system capable of executing multi-step, cross-app tasks. This shift is powered by Apple's large language models and hinges on two technical pillars: aggressive on-device processing and a redesigned framework for third-party app integration.

The architectural focus remains on privacy. Most queries will run locally on Apple's A-series and M-series chips, leveraging neural engines built over several generations. For tasks requiring more compute, Apple will route requests through its Private Cloud Compute system, a server architecture designed to prevent data access, even by Apple.

For the system to work, Apple needs developer buy-in. The new 'Siri Action Protocols' aim to replace the rigid, limited SiriKit with a standardized API contract, allowing the LLM-powered Siri to perform actions within apps, not just open them. This turns Siri from a standalone feature into an intelligent layer across the OS. Success depends on whether developers adopt this framework more widely than its predecessors.

The engineering trade-off is clear: Apple sacrifices the vast scale of cloud-based data processing used by rivals for a privacy-preserving, on-device approach. The system's capability against cloud-trained models is unproven, but the gap appears to be closing. If the technical execution matches the ambition, it could redefine user expectations for privacy and system-level AI integration.

Source: Webpronews