Apple Unveils Siri AI, Reveals App Integration Gaps

At WWDC 2026, Apple unveiled Siri AI - a ground-up rebuild of its assistant powered by Google Gemini and Apple's on-device Apple Foundational Models (AFM) family - with a beta arriving "later this year," per TechCrunch and WSJ reporting. Craig Federighi said during the keynote: "We believe privacy in AI is non-negotiable," per TechCrunch. Counterpoint estimates 450 million Apple Intelligence-compatible devices are in circulation, per WSJ. This is Tim Cook's last WWDC as CEO; John Ternus takes over September 1. A notable post-WWDC gap: The Next Web reports iOS 27's developer beta contains a hidden Extensions framework allowing users to swap ChatGPT, Claude, and Gemini inside Siri - infrastructure Apple built but chose not to announce publicly.
What happened
Apple used WWDC 2026 to launch Siri AI, a ground-up rebuild of its assistant, and said a beta will arrive "later this year," per WSJ and TechCrunch reporting. TechCrunch confirms Siri AI is developed in collaboration with Google: "Apple said it collaborated with Google and the Gemini family of models to develop the next generation of Apple Foundation Models." The WSJ reports Counterpoint estimates Apple has shipped 450 million Apple Intelligence-compatible devices. BBC and TechCrunch coverage confirm this is Tim Cook's last WWDC as CEO; John Ternus will replace him on September 1.
Technical details (reported)
PCMag reports Apple described a family of Apple Foundational Models (AFM), including AFM Core and AFM Core Advanced for on-device workloads and AFM Cloud for server-side processing, developed with Google Gemini. PCMag further reports the system combines device context, an Apple-managed world-knowledge service, and multimodal image understanding. Craig Federighi said during the keynote: "We believe privacy in AI is non-negotiable," per TechCrunch, adding that "data is only used to execute your request, and outside experts can continue to verify this promise at any time." Post-keynote, Federighi told Tom's Guide: "We see Siri not as a separate chatbot, an unintegrated place you go and chitchat, but rather as an integral but conversational tool that you use in the moment."
App integration gaps
A notable omission from the keynote: The Next Web reports that iOS 27's developer beta contains an Extensions framework that would allow users to swap between ChatGPT, Anthropic Claude, and Google Gemini directly inside Siri - infrastructure Apple built but did not announce publicly. AppleInsider's post-WWDC retrospective (June 14) flags this and other developer tooling gaps as the key blind spots, noting that most developers are still waiting for robust APIs to leverage Apple's AI capabilities inside third-party apps.
Industry context
WSJ and BBC coverage emphasize broad device reach (Counterpoint 450M estimate) as Apple's distribution advantage while flagging developer adoption and app-level integration as the critical path to delivering on the assistant's potential. TechCrunch's editorial summary characterized WWDC as Apple "catching up" in AI - leading with fixes before features.
Business and market signals
Investing.com commentary (single-source; market opinion) frames Siri AI as a potential iCloud+ monetization lever and reports Apple shares fell nearly 2% on June 8, 2026, amid investor questions about launch timing. PCMag and WSJ document Apple's emphasis on its world-knowledge service for current-events queries.
What to watch
Practitioners should monitor:
- •whether Apple opens the hidden Extensions framework to developers
- •latency and on-device performance of AFM Core models versus Gemini cloud fallbacks
- •the scope and limits of Apple's world-knowledge service for timely information
Scoring Rationale
Apple's Siri AI launch at WWDC 2026 is a major platform event affecting 450M+ compatible devices, with confirmed Gemini integration and on-device AFM models. The hidden Extensions framework and developer tooling gaps add notable practitioner relevance. Notable for ML and mobile practitioners but not a frontier model breakthrough.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
