Apple’s WWDC keynotes have long followed a familiar arc, with new platform updates, ecosystem refinements, and the occasional hardware surprise. At WWDC 2026, however, the centre of gravity shifted. This was Apple’s clearest acknowledgement yet that artificial intelligence is no longer an add-on to its platforms. It is becoming the platform itself. 

At the heart of this shift is what Apple is now calling “Siri AI”, a reimagined digital assistant built on top of a redesigned Apple Intelligence architecture. Unlike the company’s previous tentative steps into generative AI, this iteration signals a more structural overhaul. Apple is moving toward a system where intelligence is embedded across apps, interfaces, and workflows rather than being limited to standalone features. 

That shift also raises a more difficult question. Is Apple finally catching up to rivals that have spent the past two years aggressively pushing AI into their platforms, or is it trying to define a different approach altogether?

What Apple announced: Siri AI and the next phase of Apple Intelligence

Apple’s AI strategy this year is not built around a single headline feature. Instead, it is structured as a layered system that places Siri at the centre of a broader intelligence layer across devices. 

The biggest change is the transformation of Siri itself. The assistant is now designed to be conversational, capable of maintaining context across multiple prompts, and able to execute tasks across apps without requiring users to switch manually.

  This is driven by deeper integration with Apple Intelligence. Siri can now draw simultaneously from personal context such as messages, emails, and photos, understand what is on the screen, and access broader world knowledge from the web. The result is an assistant that moves beyond command-based interactions toward intent-based responses. 

Apple is also rethinking how users interact with Siri. A new “Ask Siri” interface expands responses into a full-screen conversational view, while a dedicated Siri app allows users to revisit past queries and continue conversations across devices. This shifts Siri from a reactive tool into a more persistent interface. 

Beyond Siri, Apple Intelligence is now embedded across the system. Writing tools can generate and refine text within apps, Image Playground enables more advanced image creation and editing, and system apps such as Safari, Photos, and Messages are increasingly driven by AI-powered suggestions and automation. 

Underneath these features is a notable architectural shift. Apple confirmed that its foundation models are developed in collaboration with Google’s Gemini models, alongside its own on-device processing and Private Cloud Compute infrastructure. 

This marks a departure from Apple’s traditional approach of building everything in-house. Instead, the company is selectively integrating external AI capabilities while attempting to retain control over how those capabilities are delivered to users.

How Apple’s new AI layer is structured

Until now, Apple Intelligence has felt like a collection of features. However, this year’s version is clearly an architecture. Apple is positioning its AI not as a single assistant or model, but as a layered system that sits across the entire operating system. 

At the foundation are Apple’s core models, made in collaboration with Google, which now combine on-device processing with server-side computation through what the company calls Private Cloud Compute. This hybrid approach is not new in the industry, but Apple is emphasising how tightly it is controlled. The company says that even when requests are processed in the cloud, user data is not stored or made accessible, not even by Apple. 

On top of this sits the first functional layer, which handles the fundamentals of AI interaction. This includes text generation, image understanding and creation, and speech recognition and synthesis. These are the building blocks that power features like writing tools, Image Playground, and voice-based interactions across the system. 

The second layer is the system-level orchestrator, which connects models to the rest of the operating system. It is responsible for pulling together personal context, accessing world knowledge, understanding what is on the screen, and enabling app-level actions. 

Personal context is central to this design. Apple Intelligence builds a live understanding of a user’s data across messages, emails, photos, and other content indexed by the system. This allows Siri to answer queries like retrieving a detail from an old conversation or surfacing information from a document without requiring users to specify where to look. 

Alongside this is access to broader world knowledge. Siri can now fetch up-to-date information from the web and combine it with personal context to generate more relevant responses. The combination of these two layers is what enables more complex, multi-step interactions that go beyond simple commands. 

On-screen understanding adds another dimension. Siri can interpret what the user is currently viewing and respond accordingly, whether that means answering questions about an image, suggesting actions based on a message, or helping complete a task within an app. 

Privacy as the defining layer

Privacy remains the central thread tying this architecture together. Apple is framing its AI strategy around the idea that deeply personalised experiences can be delivered without exposing user data. 

The company’s approach relies on a mix of on-device processing and tightly controlled cloud execution through Private Cloud Compute. Apple says that even when requests are handled in the cloud, user data is neither stored nor made accessible. 

At the same time, this approach is not entirely unique. Other companies are moving in a similar direction, though often in more limited contexts. Meta, for example, has introduced Private Processing for WhatsApp, which allows certain AI features such as message summarisation and writing assistance to run in a protected cloud environment.

  However, these implementations are typically restricted to specific use cases and are not applied uniformly across all interactions with Meta AI. Apple, in contrast, is attempting to extend a similar privacy framework across its entire AI system, regardless of where the request originates. 

Whether this distinction holds in practice will depend on how consistently Apple applies these safeguards as its AI capabilities expand.

How other companies are approaching AI

Apple is not alone in moving toward a system-level AI layer. Across the industry, companies are shifting away from standalone chatbots toward AI systems that operate across apps, services, and devices. However, while the direction may be similar, the execution differs significantly.

Google

Google’s announcements during this year’s I/O conference around Gemini Intelligence signal a clear shift in how it approaches AI on Android. The company is positioning Android as an “intelligence system”, where AI is deeply embedded across apps, services, and devices rather than confined to a single interface. 

Gemini Intelligence is designed to understand on-screen context, work across multiple apps, and carry out multi-step actions with minimal user input. This includes pulling information from one app to complete tasks in another, automating workflows, and maintaining continuity across devices such as phones, laptops, and wearables. 

Google is also pushing toward more autonomous systems through Gemini Spark. Unlike a traditional assistant, Spark is designed to operate in the background, executing tasks based on schedules, conditions, and user-defined workflows. It can interact with apps, browse the web, and complete actions without constant prompts, effectively acting as a persistent AI agent. 

In that sense, Google is building a layered AI system that is structurally similar to Apple’s approach. Both companies are moving toward assistants that can understand context, work across apps, and take actions on behalf of users. 

The difference lies in how far that system is allowed to extend. 

Google’s AI layer is designed to be expansive and increasingly autonomous. It can operate across a wide range of services, including third-party apps and the open web, and is built to take initiative through proactive suggestions and background task execution. 

Apple, by contrast, is taking a more constrained approach. While Siri AI is also capable of cross-app actions and contextual understanding, it is more tightly bound to the device and the user’s personal data. Instead of pushing toward continuous background automation, Apple is focusing on interactions that remain user-driven and contained within its ecosystem. 

This creates a fundamental divergence in philosophy. Google is optimising for capability and autonomy, building an AI system that can act independently across services. Apple is optimising for control and predictability, building one that is more deliberate in how and when it acts.

Microsoft

Microsoft’s approach to AI is more segmented compared to both Apple and Google. 

On standard Windows devices, AI still largely operates at a feature level. Capabilities such as Copilot integration, summarisation tools, and content generation are embedded within specific apps and services, rather than functioning as a unified system-wide layer. 

However, this begins to change with Copilot+ PCs. Designed for devices equipped with dedicated neural processing units, the Copilot+ platform introduces a more integrated AI layer that operates across the system. Here, AI moves closer to the kind of cross-app, context-aware experience that Apple and Google are now building toward. 

Even within this structure, Microsoft’s core focus remains productivity. Its AI systems are designed to enhance workflows across tools like Word, Excel, Outlook, and Teams. 

At the same time, Microsoft is also pushing into more agentic experiences. At its Build conference this month, the company introduced Scout, an always-on AI agent integrated across Microsoft 365 services including Teams, Outlook, OneDrive, and SharePoint. 

Unlike traditional assistants, Scout is designed to operate continuously in the background, using signals from emails, calendars, chats, and documents to understand context and assist with tasks. It can prepare for meetings, manage scheduling conflicts, draft emails, and surface relevant information without requiring explicit prompts. Microsoft positions this as part of a broader category of systems it calls “autopilots”, which are designed to execute tasks on behalf of users rather than simply respond to queries. 

This reinforces Microsoft’s productivity-first approach. Even its move toward autonomous agents is grounded in workplace use cases, where context is derived from structured data such as documents, communications, and schedules.

  Looking further ahead, Microsoft is signalling an even more significant shift with Project Solara. Positioned as a platform for “agent-first devices”, Solara reimagines computing around AI agents rather than traditional applications.

  In this model, users interact with agents that interpret intent and coordinate tasks across services, while interfaces are generated dynamically based on context. Microsoft refers to this as “just-in-time UI”, where the interface adapts to the task rather than being predefined.

  This points to a longer-term direction where AI is not just a layer within the operating system, but the operating system itself.

Other models

Beyond platform owners like Google, Apple, and Microsoft, most Android smartphone makers are taking a more modular approach to AI. Rather than building full-stack AI systems from scratch, they are increasingly relying on Google’s Gemini layer as a foundation, while adding their own features and interfaces on top. 

Brands such as OPPO and OnePlus are following this model by introducing dedicated AI hubs that sit alongside the core assistant experience. These systems are designed to organise user-generated content such as screenshots, notes, and saved information into a centralised “AI Mind Space”. This layer can then provide additional context to Gemini, enabling more personalised and context-aware interactions. 

This approach allows OEMs to differentiate their user experience without having to build their own large-scale AI models. Instead, they focus on how AI is surfaced, how user data is organised, and how context is fed into the underlying assistant. 

At the same time, some companies are exploring more flexible and multi-layered AI systems. Samsung, for example, is combining Google’s Gemini-powered features with its own assistant stack. While many of its AI capabilities rely on Gemini, the company is also investing in its Bixby assistant, which is being enhanced through integrations with external AI services such as Perplexity. 

This creates a hybrid system where different AI models and assistants coexist, each handling specific tasks. Rather than relying on a single unified layer, the system distributes intelligence across multiple services, allowing for greater flexibility but also adding complexity to the user experience.

Is Apple catching up, or offering something different?

At a feature level, Apple is clearly catching up. 

Many of the capabilities introduced with Siri AI mirror what rivals already offer. AI-powered contextual suggestions in apps like Messages and Phone are similar to features such as Magic Cues on Google Pixel devices. Siri’s ability to pull context from messages, emails, and other apps to create calendar events or send replies closely resembles what Gemini Assistant can already do on Android. Even Apple’s Visual Intelligence features, which allow users to interact with on-screen or camera content, echo experiences like Gemini Live and Circle to Search. 

In that sense, Apple is not introducing entirely new categories of AI features. It is aligning itself with capabilities that have already been established across competing platforms. 

Where Apple begins to differentiate is in how some of these capabilities are implemented and where they are applied.

  One example is the integration of AI within Shortcuts. While platforms like Google’s Gemini Intelligence and Gemini Spark already offer automation by executing tasks on behalf of users, Apple is extending this further by allowing users to define their own automations using natural language. 

With Shortcuts, users can describe what they want to achieve, and the system can generate a workflow accordingly. This shifts AI from simply executing predefined or suggested tasks to enabling users to program their own logic without requiring technical knowledge. The distinction is subtle but important. Google’s approach focuses on automating tasks across apps and services, often driven by context and system intelligence. Apple’s implementation, on the other hand, gives users more direct control over how those automations are created and structured. 

Another example is the Passwords app integration. Apple is using an agentic approach to automatically navigate websites and update login credentials when passwords are changed. While agentic AI is becoming more common, this is a more targeted implementation focused on a specific, high-frequency use case rather than a broad, open-ended assistant. 

This highlights a broader pattern in Apple’s approach. While competitors like Google are building agentic systems that can operate across a wide range of services and scenarios, Apple is applying similar ideas in more contained and purpose-driven ways.

New AI features and capabilities

Siri AI and system intelligence:

• 
  A redesigned Siri AI with conversational capabilities and multi-step task execution
• 
  Ability to maintain context across multiple prompts and interactions
• 
  On-screen awareness to understand and act on visible content
• 
  Access to personal context across messages, emails, photos, and apps
• 
  Integration with Spotlight on Mac for conversational queries
• 
  Dedicated Siri app to manage conversations and history
• 
  “Ask Siri” full-screen interface for more detailed interactions
• 
  Customisable Siri voice, tone, and pacing
• 
  Visual Intelligence integrated into camera, screenshots, and system UI

Cross-app actions and automation:

• 
  Siri can perform actions across apps such as sending messages, creating calendar events, and editing content
• 
  Natural language-based automation through Shortcuts
• 
  Ability to create and modify workflows using AI prompts
• 
  App Actions framework enabling deeper third-party app integration

Writing and communication tools:

• 
  AI-powered writing assistance across system and third-party apps
• 
  Automatic proofreading and tone adjustments
• 
  Contextual suggestions in Messages and Phone apps
• 
  Smart replies and content generation based on user context

Image generation and editing:

• 
  Image Playground with photorealistic image generation
• 
  Ability to use reference images from Photos
• 
  Consistent subject generation across multiple images
• 
  Tools to edit generated images by selecting specific areas

Photos app features such as:

• 
  Extend to expand images beyond original frame
• 
  Clean Up tool for removing unwanted objects
• 
  Spatial Reframing to adjust perspective and composition

Productivity and system features:

• 
  Passwords app with agentic AI to automatically update and manage logins across websites
• 
  Calendar and event creation using natural language inputs
• 
  Contextual suggestions in Phone app during calls or interactions

Safari enhancements including:

• 
  AI-based tab organisation by topic
• 
  “Notify Me” feature to track webpage updates
• 
  Ability to create custom extensions using natural language

Search and system intelligence:

• 
  Rebuilt system-wide search index for better context awareness
• 
  Faster and more accurate search across files, emails, and photos
• 
  Improved ranking and relevance in Mail and Spotlight

Other integrations:

• 
  Apple Intelligence in Maps with enhanced visual rendering and detail
• 
  AI-powered suggestions across system apps
• 
  Integration with AirPods and CarPlay for extended Siri interactions
• 
  Support for Apple Vision Pro with spatial AI interactions

Developer and ecosystem support:

• 
  New AI framework allowing developers to integrate Apple Intelligence into apps
• 
  Support for third-party AI models such as Google Gemini within apps
• 
  Expanded app-level access to system intelligence features

Availability and rollout

The next generation of Apple Intelligence, including Siri AI, is available for developer testing starting June 8 through the Apple Developer Program. A public beta will follow next month through the Apple Beta Software Program, with a broader rollout expected this fall alongside iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27.

For users, Siri AI will be released as a beta, initially limited to devices set to English, with support for additional languages expected to expand over time.

Availability will also depend on hardware. Apple Intelligence features are limited to newer devices, including iPhone 16 models and later, iPhone 15 Pro and Pro Max, iPads and Macs powered by M1 chips or newer, Apple Vision Pro, and select Apple Watch models such as Series 9, Ultra 2, and SE 3 when paired with a supported iPhone.

There are also regional restrictions. Siri AI will not be available at launch in China, and on iOS and iPadOS devices in the European Union. Apple says it is working to address regulatory challenges in these markets.

In addition, some features will have usage limitations. Tools such as image generation, which rely on server-side models, will be subject to daily limits. Expanded access will be available through certain iCloud+ subscription plans.