Siri AI at WWDC 2026

2026-06-08 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Advanced, quick

Summary

Apple's WWDC 2026 introduced new Siri AI features and a Core AI library, building on the 2024 Apple Intelligence announcements. The Siri AI, which looks feasible with current technology, licenses a custom Gemini-derived model and runs on Apple's Private Cloud Compute. It leverages vision LLMs to extract screen information, bypassing the need for app-specific integration. The Core AI library enables developers to utilize Apple hardware for their own models, integrating with Meta's PyTorch ecosystem via `coreai-torch` extensions. While an iOS 27 Developer Beta is available, Siri AI access requires a waitlist. Notably, Private Cloud Compute for demanding tasks now extends to Google Cloud systems using NVIDIA GPUs, maintaining Apple's security and privacy protections through layered architectural patterns and published binaries.

Key takeaway

For AI Engineers evaluating Apple's ecosystem for model deployment, the Core AI library's PyTorch integration signals a robust platform for on-device inference. You should investigate the `coreai-torch` extensions to leverage Apple hardware. Furthermore, the expansion of Private Cloud Compute to Google Cloud with NVIDIA GPUs, while maintaining security, offers a scalable option for more demanding agentic tasks, potentially influencing your cloud strategy for Apple-centric applications.

Key insights

Apple's new AI capabilities integrate vision LLMs and external cloud compute while prioritizing privacy and developer access.

Principles

• Vision LLMs can extract screen information without app-specific integration.
• Layered security patterns can extend private cloud compute to external infrastructure.

Method

Core AI PyTorch Extensions (`coreai-torch`) bridge PyTorch models (exported as a `torch.export.ExportedProgram`) to a Core AI `AIProgram` by traversing the FX graph and mapping ATen operators.

In practice

• Developers can use Core AI library to run PyTorch models on Apple hardware.
• Siri AI leverages vision LLMs for context-aware interactions.

Topics

• Siri AI
• Apple Intelligence
• Vision LLMs
• Private Cloud Compute
• Core AI
• PyTorch
• NVIDIA GPUs

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Engineer, Machine Learning Engineer, Tech Journalist

Related on AIssential

• OpenAI's Latest GPT 5.5 Features — Major tech companies are rapidly integrating advanced AI models while navigating significant privacy and revenue challenges.
• I Turned My M1 MacBook Into an Offline AI Coding Agent — $0 API Cost, Zero Cloud — Capable AI coding agents can run entirely offline on consumer Apple Silicon hardware, eliminating cloud dependencies.
• What to expect from WWDC 2026: Siri’s highly anticipated revamp and Apple Intelligence updates — Apple is deeply integrating advanced AI, including Google's Gemini, across its core products and services at WWDC 2026.
• Siri’s AI Comeback Could Run Through Google and Nvidia — Apple's Siri will integrate Google Gemini and Nvidia Blackwell via a hybrid on-device/cloud architecture for advanced AI capabilities.
• Apple to Renew Push for AI That Runs on Devices, Instead of the Cloud — Apple prioritizes on-device AI for privacy, speed, and independence from cloud infrastructure.
• not much happened today — AI progress is driven by aggressive model scaling, local inference optimization, and maturing agentic workflows.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.