At WWDC 2026, Apple took the wraps off the AI Intelligence layer powering a completely rebuilt Siri — marking the most significant overhaul of Apple's virtual assistant since its original launch.
What's New
The rebuilt Siri introduces several capabilities that fundamentally change the assistant paradigm. Rather than responding to individual commands, the new Siri understands context across conversations, apps, and time. It can manage multi-step workflows across multiple apps, maintain awareness of ongoing tasks, and proactively surface relevant information.
The AI Intelligence layer runs primarily on-device using Apple's custom silicon, with cloud processing handled through Private Cloud Compute — Apple's infrastructure for secure, privacy-preserving AI inference.
Technical Architecture
Apple's approach to AI is distinctive. While competitors focus on cloud-scale models with maximum capability, Apple prioritizes efficient on-device models that provide good-enough performance while maintaining strict privacy guarantees.
The AI Intelligence layer uses a combination of small on-device language models, specialized task models for specific functions, and a coordination layer that routes tasks to the appropriate model based on complexity and privacy requirements.
Implications
Apple's approach represents a fundamentally different vision from the cloud-centric models of OpenAI, Google, and Microsoft. By prioritizing on-device processing, Apple sacrifices some capability for privacy and performance — a trade-off that may prove prescient as regulatory scrutiny of cloud AI processing intensifies.
