Apple’s Rebuilt Siri Opens Its Platform to AI Rivals at WWDC 2026

Tim Cook stepped onto the Apple Park stage this morning for his final WWDC keynote as chief executive, and the product at the center is the one his tenure failed to deliver: a Siri that works as a real AI assistant. Apple’s 37th Worldwide Developers Conference, running June 8 through June 12, puts a rebuilt Siri at its center, two years after the company first promised the upgrade and one month after agreeing to pay $250 million to settle a false-advertising lawsuit over those same features.

The chat-mode Siri and personal-context unlock have absorbed most of the pre-event coverage. Underneath those consumer features is an Extensions system that gives rival AI models, Claude, Gemini, and ChatGPT, their first direct access to the iPhone’s system layer.

What the New Siri Does

The rebuilt assistant ships as a standalone app, the first time Siri has had one since it launched on the iPhone 4s in 2011. Instead of the familiar glowing orb, the new Siri lives in the Dynamic Island. A swipe down from the top center of the screen opens a “Search or Ask” panel that replaces Spotlight as the primary search gesture across iOS. Chat history persists across sessions in an iMessage-style conversation view, with auto-delete options at 30 days, one year, or indefinitely.

On capability, iOS 27 delivers what Apple demonstrated at WWDC 2024 but never shipped:

• Personal context: Siri reads emails, photos, messages, calendar entries, contacts, and files to inform its answers
• On-screen awareness: the assistant sees and acts on whatever is visible in any open app
• Multi-step task execution: Siri chains actions across apps without the user opening each one
• Web queries: answers from the internet surface inside the same chat interface
• Document uploads: users can submit files directly in the chat view for Siri to process

At launch, the most demanding Apple Intelligence features are expected to require at least an iPhone 15 Pro. Full support extends across the iPhone 17 lineup when iOS 27 ships publicly in September.

Google Powers the Engine

The intelligence upgrade runs on a model Apple doesn’t own. In January 2026, Apple and Google published a joint statement confirming that the next generation of Apple Foundation Models would be built on Google’s Gemini technology, processed through Apple’s Private Cloud Compute (PCC) infrastructure rather than Google’s own servers. Bloomberg’s Mark Gurman reported the deal at roughly $1 billion per year. Thomas Kurian, Google Cloud’s chief executive, reaffirmed the scope at Google Cloud Next ’26, saying Apple and Google are “collaborating as Apple’s preferred cloud provider to develop the next generation of Apple Foundation Models based on Gemini technology.”

• 1.2 trillion parameters in the licensed Gemini model Apple routes through PCC, roughly eight times the parameter count of Apple’s largest prior cloud model
• ~3 billion parameters in Apple’s on-device model, handling privacy-sensitive queries locally
• ~$1 billion per year: the reported annual licensing cost to Google
• 0 queries retained: PCC uses stateless, ephemeral processing with personally identifiable information stripped before the Gemini inference layer

The privacy arrangement carries an unusual contractual term: Apple’s agreement prevents Google from using Siri queries to train future Gemini models. Queries exceeding the on-device model’s capacity route to Apple Silicon servers in PCC, bypassing Google’s cloud entirely. An ACM (Association for Computing Machinery) conference paper presented this month confirmed Apple’s three core PCC privacy claims after independent analysis, per the full technical specifications of the Gemini licensing arrangement. The custom Gemini model uses a mixture-of-experts architecture that activates only a relevant subset of its parameters per query, which is how Apple keeps latency competitive on a model that large.

Apple turned to Google partly because its own AI models reportedly failed to reach the performance targets needed for the personal-context features it had been promoting. In the year before this conference, Apple shares gained roughly 50%, solid but well behind Alphabet’s roughly 120% climb over the same stretch, gains that tracked directly with Gemini’s commercial success.

The Extensions Marketplace

iOS 27 introduces an Extensions system that lets users designate any qualifying AI service as the default provider for system-wide Siri functions including Writing Tools and Image Playground. OpenAI’s exclusive arrangement as Siri’s sole external AI fallback, in place since iOS 18.2 shipped in December 2024, ends with this release. Claude, Gemini, and ChatGPT all qualify from day one, and any AI provider can join through the App Store.

The mechanism mirrors Apple’s approach to default browsers and email clients since iOS 14: a user installs their preferred AI app from the App Store, activates it as an Extension in Settings, and the surface-level intelligence layer becomes a competitive market. Apple collects App Store revenue on every qualifying AI app that joins. Andrew Cornwall, a senior analyst at Forrester, said before the keynote that Apple was expected to let developers “plug their apps into Siri using what Apple calls ‘extensions’ and let those developers choose among AI models from OpenAI, Anthropic, and Google’s Gemini in their apps.”

Apple is not, at least at this conference, extending AI access to full-screen automation. Ben Bajarin, CEO of tech consultancy Creative Strategies, said before the conference that Apple was not expected to allow the kind of screen-control AI automation that competing Android platforms have started shipping. “It’s way too early for the consumer,” Bajarin said. “Honestly, I’m not even sure businesses are ready for this in an uncontrolled context.”

The Developer Platform Shift

Two framework changes matter more to developers than any consumer-facing announcement. App Intents, available since iOS 16, lets developers declare actions Siri can trigger directly without requiring the user to open the app. Apple is expected today to add “App Intent Domains,” structured action categories covering messaging, documents, media, and productivity, which let AI agents chain cross-app task sequences from a declared intent library. Apple’s developer documentation for integrating App Intents with Siri and Apple Intelligence already covers how actions conform to assistant schemas; the schema set is expected to expand substantially with today’s release.

Potentially larger in impact over time: an expected expansion of the Foundation Models framework to multimodal (image) input. The framework shipped with iOS 26, giving apps direct Swift access to the on-device model for text tasks. Developers handling image analysis currently bundle their own vision models at 1 to 2 gigabytes per app. If Apple extends Foundation Models to images today, those bundled weights become replaceable with a single system API call, cutting binary sizes and letting model quality improve automatically with each OS update.

Patrick Moorhead, founder of tech consulting firm Moor Insights & Strategy, framed the platform question before the conference: “They have to make Siri not suck, but Apple also has to put the framework together of how their developers can take advantage of AI themselves.” The developer beta dropping today will contain the API documentation that answers it.

A $250 Million Legal Reckoning

WWDC 2026 opened under a legal cloud Apple built for itself. Apple previewed the personal-context and on-screen awareness Siri features at WWDC 2024, promoted them in advertising during the iPhone 16 launch in September 2024, then delayed them in March 2025 and pulled the ads.

We resolved this matter to stay focused on doing what we do best, delivering the most innovative products and services to our users.

That was Apple’s statement in May 2026 when it agreed to pay $250 million to settle the class-action lawsuit Landsheft v. Apple Inc. (Case No. 5:25-cv-02668-NW) in the U.S. District Court for the Northern District of California. The suit alleged Apple deceived consumers with a marketing campaign promoting Siri capabilities “that did not exist or were materially misrepresented.” Apple denied wrongdoing. Court filings estimate roughly 36 million eligible devices fall within the settlement class: iPhone 15 Pro, 15 Pro Max, and every iPhone 16 model purchased in the U.S. between June 10, 2024, and March 29, 2025. Eligible buyers can claim $25 per device, rising to $95 if claim volume is low. The preliminary approval hearing sits before Judge Noël Wise on June 17.

Cook presented today alongside John Ternus, the hardware engineering chief who takes the CEO title on September 1 in a succession Cook announced in April. Ternus’s first major product moment as CEO will be the iPhone 17 launch in the fall, the device the rebuilt Siri is designed to run across. Gurman has written that even a convincing WWDC demo leaves the question of whether Siri can undo years of brand damage in everyday use. The settlement is the record on these specific features.

Frequently Asked Questions

Who Qualifies for the Apple Intelligence False Advertising Settlement?

U.S. customers who purchased an iPhone 15 Pro, iPhone 15 Pro Max, or any iPhone 16 model between June 10, 2024, and March 29, 2025 qualify. Both individuals and businesses are included in the settlement class. The case, Landsheft v. Apple Inc. (Case No. 5:25-cv-02668-NW), is pending before Judge Noël Wise in the U.S. District Court for the Northern District of California. Court filings estimate approximately 36 million eligible devices. Full eligibility details are in the Apple Intelligence false advertising settlement summary.

How Much Can Eligible Buyers Claim, and When?

The presumptive payout is $25 per eligible device, rising to as much as $95 per device if overall claim volume is low. Buyers with multiple eligible devices can file separately for each. The $250 million fund is non-reversionary. Claim submission instructions will appear on the settlement administrator’s website within 45 days of the May 5, 2026 preliminary approval filing, pending the court’s decision at the June 17 hearing.

Does the Rebuilt Siri Require an iPhone 16?

No. Apple plans to support the rebuilt Siri’s personal-context and chatbot features on iPhone 15 Pro and iPhone 15 Pro Max at launch alongside iPhone 16 models. Full support extends across the iPhone 17 lineup when iOS 27 ships publicly in September. Older devices without a compatible Neural Engine will not receive every Apple Intelligence feature even with the software update.

What Is Apple’s Extensions Framework for Siri?

Extensions is a system in iOS 27 that lets users set a third-party AI service as the default provider for Siri system-wide functions including Writing Tools and Image Playground. Qualifying apps including Anthropic Claude, Google Gemini, and OpenAI ChatGPT install from the App Store and activate in Settings under Apple Intelligence and Siri. This ends OpenAI’s position as the only external AI option inside Siri, which has been in place since iOS 18.2 shipped in December 2024.

When Does iOS 27 Ship to the Public?

Developer betas drop June 8, the same day as the WWDC keynote. A public beta follows in July. The full release, alongside new iPhone hardware, is scheduled for September 2026. Some Apple Intelligence features may ship with a beta label even in the September release.

The settlement’s preliminary approval hearing is on June 17, nine days away, covering the exact capabilities Apple demonstrated on stage today.