WWDC 2026 Keynote

Highlight

At WWDC26, Apple announced iOS 27, iPadOS 27, macOS Golden Gate, watchOS 27, tvOS 27, and visionOS 27. The main upgrades revolve around three themes: platform responsiveness and refinements to the Liquid Glass design, safety and privacy tools for children, and Siri AI plus the next generation of Apple Intelligence built on an Apple-Google Foundation Models architecture.

Core Content

System-level refinement

Opening an app, switching pages, or transferring files used to feel just a little short of instant. In iOS 27, Apple preloads key data so iPhone and iPad apps launch 30% faster. Developers do not need to change any code; the system handles it automatically. (10:28)

After taking a photo, new photos now appear in the Photos app 70% faster. AirDrop transfers are 80% faster. File transfers from an external drive connected to an iPad are 5x faster, matching Finder speeds on the Mac. (11:07)

More surprisingly, the iPhone 11 can also run iOS 27. Apple optimized the advanced CPU scheduler that had previously been limited to newer models and brought it to iPhone 11 and later, keeping a phone from 2019 smooth in 2026. This is the iOS release that covers the largest user base ever. (12:01)

Network switching is smarter too. In the past, when you reached the doorway of a coffee shop, your phone might cling to a weak Wi-Fi signal until you manually switched to cellular. Now iPhone can decide when to stay and when to move on. When you send a large file on a low-bandwidth connection, Messages shows a sending progress indicator instead of leaving you waiting. (12:33)

Search is another part of the system that has been completely rewritten. The search indexes for Spotlight, Photos, and Mail have been re-architected to be more stable and more efficient. After updating, the system automatically rebuilds its indexes, and new content is indexed almost in real time. Mail also introduces a new ranking system so the message you want is more likely to appear first in Top Hits. (13:41)

Liquid Glass design iteration

The Liquid Glass design language introduced last year continues to be refined this year. (06:46) Apple improved the way Liquid Glass diffuses complex background content, making text and controls easier to read. A new transparency slider, from ultra clear to fully tinted, lets users adjust the effect themselves. Apps that have already adopted Liquid Glass automatically support this customization.

On macOS, toolbars return to a more unified design, sidebars extend to the window edge, icons regain color, and window corner radii are more consistent. App icons themselves now incorporate more Liquid Glass layering, with refraction effects that make them look sharper. (08:01)

Child safety and parental controls

Parental controls became a major focus this year. The core idea is that every child is different and parents know their own children best, but the system should provide default recommendations based on expert research. (18:00)

New features include:

• Child Account: New child accounts automatically enable protections based on age, and existing accounts can be converted. For children under 13, Ask to Buy and the new Ask to Browse are enabled by default. Ask to Browse requires parental approval before visiting new websites, and both flows are handled through Messages. (19:54)
• Time Allowances: Parents can set daily usage time by category, including Entertainment, Games, and Social Media. Recommended values are based on age and research from the American Academy of Pediatrics. Social media is disabled by default for children under 13. (23:53)
• Schedules: Parents can control which apps are available by time period, such as allowing only learning apps during class time. Combined with Time Allowances, weekends can get extra time for movies. (25:07)
• Expanded Communication Safety: In addition to nudity, the system now intercepts graphic violent images and videos, including during live FaceTime calls. (23:14)
• Redesigned Screen Time: Parents can see a child’s device usage at a glance and adjust access with one tap. (25:43)

For developers, Apple provides the Declared Age Range API, which lets apps learn a user’s age range while preserving privacy and then adjust content presentation accordingly. Communication Safety API and Contact Approval API also help build safer experiences for children. (26:54)

The next generation of Apple Intelligence and Siri AI

This is the biggest architectural change this year. Apple has deeply collaborated with Google to build the next generation of Apple Foundation Models based on Gemini technology, with support for both on-device execution and Private Cloud Compute servers. (29:29)

The new architecture contains four core capability layers:

1. Personal context understanding: Understands your photos, notes, mail, and other content through Spotlight’s semantic index. Any app that integrates with Spotlight can participate. (30:57)
2. Broad world knowledge: Retrieves up-to-date information from the web and generates answers through Private Cloud Compute. (31:17)
3. App Actions: Finds the right tool from an app’s toolbox to complete a request, such as drafting an email or editing a photo. (31:31)
4. On-screen awareness: Offers targeted help based on the app you are using and what you are doing in it. (31:45)

All of these capabilities are coordinated by the new System Orchestrator. On privacy, Apple emphasizes that AI privacy is non-negotiable: processing happens on device plus Private Cloud Compute, data is not stored, and neither Apple nor third parties can access it. (32:17)

Siri has been completely rebuilt. The new version is called Siri AI. It supports multi-turn conversations, on-screen content recognition, personal context, and a new Siri App for reviewing conversation history. History syncs privately through iCloud so conversations can continue across devices. (33:45)

Siri AI also brings Visual Intelligence. In the iPhone Camera app, a new Siri mode can look at food and provide nutrition information, or look at a bill and split the payment. On macOS, it is triggered with a keyboard shortcut, letting you select screen content and ask a question directly. On iPad, it is integrated into the screenshot experience. On visionOS, you can look directly at an object and ask Siri without saying “Hey Siri.” (47:20)

Writing Tools have also been upgraded. They can now draft documents from scratch, imitate your communication style with a specific contact in Mail and Messages, and automatically proofread. Automatic proofreading is available system-wide, including in most third-party apps. (49:41)

Details

Siri AI in real use

In the demo, Mike showed how Siri AI connects multiple capabilities in everyday scenarios:

Scenario 1: Combining world knowledge and personal context

The user asks, “When is Suki Waterhouse playing in San Francisco?” Siri retrieves the information from the web and answers. The user then asks, “How do I get tickets?” and Siri explains that tickets require a lottery. The user says, “Remind me when the lottery opens,” and Siri creates a reminder. Finally, the user says, “Play her new song,” and Siri opens Music and starts playback. The whole conversation flows naturally. (35:06)

Scenario 2: On-screen awareness plus personal context

The user sees a beach photo and asks, “Where is this?” Siri identifies it as Natural Bridges State Beach in Santa Cruz. The user remembers that a friend recently moved nearby and asks, “Where is Jeff’s new house?” Siri finds the address in Messages. Then the user says, “Navigate me to that arch, stopping at Jeff’s house along the way,” and Siri automatically plans the route. (36:05)

Scenario 3: Multi-turn creative conversation

The user asks, “What is the World Cup opening weekend schedule?” Siri provides the schedule. The user scrolls down into conversation mode and says, “Brazil versus Morocco would be great for a watch party. Give me representative dishes from both countries.” Siri offers menu ideas with images. The user remembers that his daughter wanted to bring dessert and asks, “What dessert did Maria mention recently?” Siri finds “coconut cookies” in Messages. Finally, the user says, “Combine these into a unique party menu,” and Siri generates a creative menu that blends dishes from both countries and can be sent directly to the group chat. (39:56)

Scenario 4: Siri on macOS

From Spotlight, the user asks, “How do I build a makerspace in a shed?” Siri gives detailed suggestions. The user selects three quote files in different formats, Control-clicks, and asks, “Compare these and help me choose one.” Siri generates a comparison table. The user remembers that his son mentioned an electrical issue and asks, “Which one can solve that?” Siri finds the relevant content in Messages and recommends an option. Then the user says, “Draft a friendly email to Modular Workshops asking whether they can speed up delivery,” and Siri automatically extracts the contractor’s email and name and writes the full message. (43:01)

AI enhancements in Safari

Safari gets three practical features this year.

Tab Topics: Apple Intelligence automatically groups open tabs by topic. As you browse, newly opened related pages join the matching topic. When you are done, you can close the whole topic with one tap or save it as a Tab Group. (52:28)

Notify Me: Tell Safari what change you are waiting for on a webpage, such as “registration opens for this camp,” then close the tab. Safari monitors it in the background and sends a notification when the change is detected. (53:28)

Describe an extension: Describe the browser extension you want in natural language, and Safari automatically generates a custom extension that changes webpage content. For example: “Add a button to the toolbar to save and rate recipes I’ve tried.” (54:13)

Automatic password repair

The Passwords app can now automatically update weak or leaked passwords to strong ones with one tap. Behind the scenes, Apple Intelligence and Safari work together to sign into each website, navigate to the password change page, and complete the update flow as an agent. (54:37)

Call Context

During a phone call, the Phone app automatically finds relevant information from Mail, Calendar, and other apps. For example, if you call an airline to change a flight, the Phone app can automatically find and display your confirmation code. The whole process runs on device. It only looks at “who you called,” not “what you said,” so neither Apple nor the airline can see it. (56:29)

Creating Shortcuts with natural language

Previously, creating Shortcuts meant manually dragging action nodes. Now you can simply describe what you want. For example: “When I leave work, message Pedro that I’m on my way and include my estimated arrival time.” Shortcuts automatically assembles the trigger, calculates the ETA, and sends the message. If you want to adjust it, keep describing the change, such as “also start playing my favorite podcast.” (59:14)

Image Playground upgrades

Image Playground now supports a photorealistic style, powered through Private Cloud Compute. It can generate images based on people in the Photos library, transform a photo’s style from a natural-language description, and let users circle objects to move, scale, or add elements. It also adds multiple sizes, including landscape for websites and portrait for flyers. It is integrated into Contact Posters, Lock Screen wallpapers, and Messages backgrounds. (01:00:30)

Developers can integrate these capabilities through the Image Playground API. (01:02:27)

AI editing in Photos

There are three new tools:

• Upgraded Clean Up: Better quality and more realistic fills, even for complex scenes. (01:03:20)
• Extend: Extends image boundaries to give the subject more space or straighten the horizon without cropping content. (01:03:41)
• Spatial Reframing: Uses an on-device spatial model for real-time preview, letting you adjust a photo’s viewpoint as if you had moved the camera. Missing edges are filled by a generative model on Private Cloud Compute, and new content is generated only where the viewpoint shifts, keeping it consistent with the original scene. (01:04:23)

These features work with almost all photos, including old photos and photos taken with other cameras. (01:06:37)

Developer tools

• Foundation Models framework: The on-device model now supports image input, custom skills, server models, and a unified Swift API. The Daydream app demonstrates the real workflow: pick a photo, have the on-device model identify clothing items, then use fashion knowledge to recommend outfits. (01:09:38)
• Core AI framework: Runs other models on every platform and uses the full power of Apple Silicon for acceleration. This brand-new framework gives third-party models the same level of hardware acceleration as Apple’s own models. (01:10:27)
• Xcode Coding Agent: Supports localizing an entire app, interacting with simulated devices, custom skills, model selection including Gemini, and connections to Figma and GitHub. (01:11:02)
• Device Hub: A unified interface for managing all simulators and physical devices, with multi-touch simulation, one-tap appearance switching, and dynamic resizing. (01:11:21)

Other highlights

• Photos shared albums: Android and Windows users can now join iCloud shared albums and contribute photos, with full-resolution sharing supported. (14:56)
• Health: Cycle Tracking adds support for perimenopause and menopause, recognizing changes in cycle patterns and recording symptoms. (15:14)
• AirPods: Custom EQ further personalizes sound quality. (15:34)
• Vision Pro: Panoramic photos can be converted into spatial scenes and used as personal environments. (15:47)
• Maps Flyover: Combining aerial imagery with visual intelligence models makes building details, tree shapes, and glass reflections render more sharply. (16:07)

Availability notes

Siri AI initially supports English only and will quickly expand to more languages. A beta arrives later this year. It is not currently available on iOS and iPadOS in the European Union, where Apple is still looking for a privacy-preserving solution. In mainland China, Siri AI and other new Apple Intelligence features are unavailable until regulatory requirements are resolved. (01:08:16)

All Apple Intelligence features support the same devices as today. The most powerful on-device models and their corresponding features, including expressive speech and advanced dictation, are available only on the most powerful iPhone, iPad, and Mac models. (01:07:54)

Core Takeaways

1. Adapt your app for Siri AI App Actions

• What to do: Make your app callable by Siri AI through natural language.
• Why it is worth doing: Siri AI can now hold multi-turn conversations, understand on-screen content, and use personal context. If your app exposes App Intents, users can say “Add the photos from Shasta last week to the family shared album” instead of manually opening Photos. The Line app indexes message content into Spotlight so users can ask Siri to find chat history directly. The Structured calendar app uses App Intents so Siri can create events directly.
• How to start: Use the AppIntents framework to define your core actions and make sure content is indexed by Spotlight. See Session 345 (App Intents New) and 343 (App Intents Advanced).

2. Integrate Image Playground API for personalized content

• What to do: Let users generate custom images in your app, such as Contact Posters, chat backgrounds, or event invitations.
• Why it is worth doing: Image Playground now supports photorealistic style and protects privacy through Private Cloud Compute. Users do not need to upload photos to third-party services.
• How to start: Use the ImagePlayground framework and pass in people from the user’s Photos library plus a text description.

3. Use Declared Age Range API for content ratings

• What to do: Dynamically adjust app content based on the user’s age range.
• Why it is worth doing: Apple made child safety a central theme this year, and parental controls will strictly limit noncompliant apps. Adopting Age Range API early helps prevent your app from being blocked in Family Sharing contexts. Social media is disabled by default for children under 13, so apps need to adjust features and content based on age.
• How to start: Call the DeclaredAgeRange API in your app and adjust content presentation and interactions based on the returned age range.

4. Adapt to Liquid Glass design

• What to do: Make sure your app presents Liquid Glass correctly on iOS 27 and macOS Golden Gate.
• Why it is worth doing: Users can now adjust Liquid Glass transparency from ultra clear to fully tinted. If your app already uses the system-provided Liquid Glass APIs, these settings take effect automatically. On macOS, sidebars extend to the edge, icons regain color, and window corner radii are unified, so even apps that have not been updated benefit.
• How to start: Check whether your UI uses system materials and toolbar APIs, and avoid custom implementations that conflict with system settings. See Session 269 (SwiftUI) and 289 (AppKit Modernization).

5. Explore Core AI and Foundation Models for local models

• What to do: Embed custom models in your app or call Apple’s Foundation Models, running fully locally or on Private Cloud Compute.
• Why it is worth doing: The Foundation Models framework now supports image input, custom skills, server models, and a unified Swift API. Core AI is a new framework that lets third-party models run locally with full Apple Silicon performance across all platforms. The Daydream app shows the practical workflow: use an on-device model to recognize clothing, then recommend outfits.
• How to start: Watch Session 241 (Foundation Models), 324 (Core AI), and 328 (MLX Swift) to learn about model conversion and invocation.

Related Sessions

• Siri AI and the new iPhone experience — A deep look at Siri AI’s iPhone-specific interaction design and Visual Intelligence
• Foundation Models — Technical details of Apple Foundation Models, including on-device and cloud model architecture and invocation
• Agentic Apps — How to build intelligent apps that can carry out tasks on their own, with a detailed look at Siri AI’s App Actions mechanism
• What’s new in App Intents — Concrete implementation techniques for making app features discoverable and callable by Siri AI
• Core AI — A new framework and best practices for running custom ML models inside apps