Alex Zaharov-Reutt, Global AI and Technology Editor | Published 9 June 2026
Apple has finally shipped the thing it spent two years promising. At WWDC26 it stripped Siri back to the studs and rebuilt it as Siri AI, a chatty, context-aware assistant that can reach into your messages, email and photos and actually get things done. The twist sits one layer down, and Apple said it plainly on stage: the brains come from a deep collaboration with Google and its Gemini models.
A Siri that finally does what the 2024 demo promised
Let's start with what was announced, because there's a lot of it. In its keynote release, Apple unveiled the next generation of Apple Intelligence and Siri AI across iOS 27, iPadOS 27, macOS 27 (named Golden Gate this year, after the now-traditional naming gag from Apple's software team), watchOS 27, tvOS 27 and visionOS 27.
Siri AI is a full teardown, right down to the foundations. It draws on what Apple calls personal context understanding to search across your messages, emails and photos, answers questions on almost any topic using live web knowledge, and reads what's on your screen so it can act on it.
The examples Apple gave are the ones people actually ask for: surface a hotel confirmation number buried in an old email, find the restaurant a friend texted you about last month, or pull up the photos from last weekend by who's in them.
There's a dedicated Siri app now, syncing your conversation history privately across devices through iCloud, so you can start on the Mac and pick it up on iPhone. You invoke it the old way with "Hey Siri" or the side button, or by swiping down from the Dynamic Island. On iPad and Mac it lives inside Spotlight, and on Apple Vision Pro it floats as a 3D blob you can park in your room and talk to.
Here is Apple's full WWDC 2026 presentation – MUST WATCH!
Visual Intelligence widens out too. A new Siri mode in the iPhone Camera lets you point, tap the shutter and ask what you're looking at, then split a restaurant bill with friends over Apple Cash or get the nutrition rundown on a plate of pho. The same visual smarts come to iPad, Mac and Vision Pro for the first time.
Apple also wired Writing Tools through the whole system, so Siri can draft an email from a one-line description, or rewrite it to match how you usually talk to a given contact. On the priciest silicon (iPhone Air, iPhone 17 Pro and Pro Max, iPad with M4, Mac with M3 and at least 12GB of memory, and Vision Pro on M5) you get more expressive, adjustable voices and noticeably sharper dictation.
Apple's full Siri AI release carries the company line.
"With access to broad world knowledge for up-to-date answers on virtually any topic, along with onscreen awareness and personal context understanding, Siri AI can help users take action across apps more naturally than ever," said Apple senior vice president of software engineering Craig Federighi, whose remit covers the company's core AI work (his official bio sits on the Apple Leadership page, as Apple's software chiefs famously don't keep public LinkedIn profiles).
The Google Gemini collaboration Apple announced from the stage
Now the part a tech reporter can't skip past, and credit where it's due, Apple didn't bury it.
From the keynote stage, Federighi put it in plain words: "This year we embarked on a deep collaboration with Google, leveraging the technologies behind their Gemini family of models. Together, we created the next generation of Apple Foundation Models." The written everyday experiences release and the developer tools announcement use the same phrase, models "custom-built in collaboration with Google and its Gemini models."
The privacy company is now building its AI on a foundation co-created with the world's biggest advertising business, and it chose to say so itself rather than leave it to the teardown crowd.
This didn't come from nowhere. Back in January, Apple and Google confirmed a multi-year partnership, widely reported at around US$1 billion a year, for a custom Gemini model to power a rebuilt Siri. TechCrunch and a rare joint Google-Apple statement sketched the shape of it at the time.
The Google fingerprints don't stop at the model. Every photo you edit or generate with Apple Intelligence now gets a hidden SynthID watermark, which is Google DeepMind's provenance technology, stitched straight into Photos and Image Playground.
So here's the picture, plainly: the foundation underneath Siri AI was co-built with Google's Gemini, then adapted by Apple to run on device and on Private Cloud Compute. Apple has spent a decade selling silicon and software it controls top to bottom. On the biggest software bet of the decade, it went to Google for the engine.
Rebuilt architecture, with privacy as the entire sales pitch
Apple's defence against that uneasy read is its architecture, and to be fair it's a serious piece of engineering. Siri AI runs on a mix of on-device models and Apple's Private Cloud Compute, which Apple says keeps your personal data off its servers and out of anyone else's hands, with outside experts free to verify the claim.
A new "system orchestrator" decides what runs locally versus in the cloud, while the Spotlight index and what Apple calls App Toolbox stay entirely on device. Apple's pitch, repeated like a mantra, is that Siri "remains the world's most private digital assistant."
On stage, Federighi took a swing at rivals while he was at it. "Today many AI providers talk about privacy, but by default most of them retain your personal interactions," he said, before calling privacy in AI "non-negotiable." It's a confident thing to say about an assistant whose foundation models were co-built with Google.
That's the tension in one sentence. The model is co-built with Google, the guard rails are Apple's, and whether that combination holds up to scrutiny is the question I'll be watching all year – as will the rest of the tech journalists and tech industry.
Apple Intelligence grows up across the apps you actually use
Beyond Siri, the everyday wins are the ones that'll land with normal people.
Photos gets genuinely clever editing: Spatial Reframing lets you shift the camera angle after the shot, Extend fills in a cramped frame, and Clean Up does a far tidier job of erasing photobombers. Safari can herd your runaway tabs into topics, watch a web page and ping you with Notify Me when a price drops or stock returns, and even build a custom extension from a plain-English description.
Passwords now fixes weak or breached logins for you, with Apple Intelligence and Safari quietly signing in and upgrading the account on your behalf – Federighi even said it was being done “agentically”. Image Playground can finally produce photorealistic images, running on Private Cloud Compute.
There's more threaded through the system: one-tap suggestions in Messages and Mail that pick up your writing style, Call Context that surfaces your booking reference when you ring an airline (processed on device) – shades of Google’s Magic Cue there, the “describe-it” Calendar events, describe-it Shortcuts, and a Home app that bundles related notifications and writes descriptions of your security-camera clips so you can search them.
Under the hood, Apple says iPhone and iPad apps launch up to 30% faster, photos load up to 70% faster after capture, and AirDrop moves up to 80% quicker. The Liquid Glass design from last year gets a transparency slider, after plenty of grumbling about legibility.
For developers: Xcode 27 goes agentic, and opens up to Claude, Gemini and GPT
This is where iTWire's developer readers should lean in, and it's the most quietly significant shift of the lot.
Xcode 27 turns into an agentic coding environment that plugs directly into models from Anthropic, Google and OpenAI. Apple names Claude and Gemini outright, and Federighi singled out Gemini support from the stage. Agents can plan, hold a back-and-forth, write and run their own tests, try ideas in Playgrounds and drive the simulator through a new Device Hub, so they can run on their own for longer. Plug-ins arrive via the Model Context Protocol and the Agent Client Protocol, with GitHub and Figma first out of the gate.
For Apple, an ecosystem that guarded its tooling like a state secret, shipping an IDE built around third-party coding agents is a real change of posture.
The Foundation Models framework grows into a single native Swift API supporting on-device models with image input, server models and custom skills. Indie developers in the App Store Small Business Program, with fewer than 2 million lifetime downloads, get access to Apple's next-generation Foundation Models on Private Cloud Compute at no cloud API cost. A new framework called Core AI lets developers run full-scale large language models locally on Apple silicon.
"Developers are at the heart of the Apple ecosystem, and our goal is to provide them with the best possible tools and technologies to build the future," said Apple vice president of worldwide developer relations Susan Prescott (one Apple executive who does keep a public LinkedIn profile). "With new intelligence frameworks and agentic coding in Xcode 27, developers have the tools they need to focus on what they do best: bringing their incredible ideas to life."
Xcode itself is now Apple silicon only and 30% smaller, Xcode Cloud is up to twice as fast, and there's a stack of new kit for game and spatial developers, including Game Porting Toolkit 4, official Unity plug-ins, a Steam Asset Converter and Reality Composer Pro 3.
The App Store gets new ways to make, and keep, money
Apple is also reworking the App Store's commercial machinery.
Developers get Creative Assets and a shared Asset Library for richer product pages, plus Personalized Collections and App Notes that explain why an app is being recommended (rolling out this week, English, in the US first). Subscriptions stretch to groups and organisations through StoreKit 2, with volume purchasing via Apple Business and School Manager landing this fall and group purchases this winter.
There are cross-developer App Store Bundles, multi-app Suites, and Retention Messaging so developers can make a last pitch when you hit cancel. The Mac App Store drops its Intel requirement, letting developers ship Apple silicon-only builds.
Child safety: Apple's answer to the under-16 era
The release that'll resonate hardest in Australia is the one about kids.
Apple previewed a new suite of child safety features built around child accounts, required for under-13s and available up to 18. Setup Assistant lets parents pick exactly which apps a child starts with, Ask to Buy covers downloads (free ones included), and the new Ask to Browse makes kids request permission before opening a new website in Safari, across iPhone, iPad and Mac.
Communication Safety, which already blurs nudity by default for under-18s, now also steps in to hide gore and violent imagery. Time Allowances let parents cap time by category (Entertainment, Games, Social Media) with age-based starting points drawn from clinical research, Schedules lock apps to certain hours, and Screen Time has been redesigned around an at-a-glance dashboard. Apple is adapting the American Academy of Pediaatrics' Family Media Plan into the experience.
"Our approach to helping families create safer digital experiences is grounded in the belief that every child is unique," said Apple vice president of health and fitness Dr Sumbul Desai, a Stanford clinical professor whose team also built the ECG and sleep apnoea features (her profile sits with the Stanford Center for Digital Health).
The developer hooks matter just as much for an Australian audience. Apple is shipping a Declared Age Range API that hands apps a child's age band without ever sharing a birthday, alongside SensitiveContentAnalysis and PermissionKit. With Australia's world-first under-16 social media restrictions now in force, an OS-level, privacy-preserving way to signal age is exactly the kind of plumbing local platforms and regulators have been arguing about. Apple just put a version of it in the box.
Europe gets cut off, and Apple isn't being subtle about why
Here's the fight. Due to the Digital Markets Act, Apple says it can't ship Siri AI in the EU on iOS 27 and iPadOS 27 at launch. EU users will get Siri AI on macOS 27, visionOS 27 and watchOS 27, but not on their iPhones or iPads, with no timeline offered.
Apple's argument is blunt. It says the regulators' reading of the DMA would force it to give any rival assistant near-unlimited, autonomous access to a user's device the moment Siri AI goes live: reading and sending messages, making purchases, touching files and acting across any app, without ongoing user oversight. It points to security researchers who've shown AI systems can be hijacked to lift passwords and photos.
Apple says it offered a fix, a "Trusted System Agent" intermediary plus an 18-month phased rollout, and that the European Commission rejected every proposal.
"We're deeply disappointed that our EU users won't have Siri AI on iPhone or iPad when we share our new software releases later this year," said Apple senior vice president of software engineering Craig Federighi.
Worth keeping both eyes open here. This sits inside a much bigger brawl: Apple has appealed the Commission's interoperability ruling and argues the rules let "data-hungry companies" like Meta and Samsung at user data. Critics, including the Free Software Foundation Europe, counter that Apple keeps reaching for privacy as a reason to avoid opening up, and note that of dozens of formal interoperability requests, almost none have produced a solution. A key DMA interoperability deadline fell on 1 June 2026. Read this delay as one chapter in that bigger standoff. China, separately, also misses out while Apple works through local rules.
Where this leaves Apple in the assistant wars
Step back and the competitive irony is hard to miss.
Google's Gemini now powers Apple's Siri AI, Samsung's Galaxy AI, and Google's own Pixel and Home gear. Amazon's revamped, generative Alexa+ is the other big player, and the early head-to-heads (Tom's Guide ran Gemini against Alexa+; The Ambient pitched them for the smart home) give Gemini the edge on conversation and Alexa the edge on running your house.
Apple's wager is about integration over raw model supremacy: the best-connected assistant that can touch your real data safely, with privacy as the differentiator. It's willing to license Gemini to get there, and to wear the awkwardness of running a rival's brain to do it. Whether customers reward the integration or just notice that Siri finally works is the open question.
What it means for Australia
Apple chief executive Tim Cook closed the keynote with the timeline: developer betas are live now at developer.apple.com, a public beta lands next month through beta.apple.com, and the free updates ship "this fall," which for us down here means spring. Siri AI itself arrives as an English-only beta later in the year, so Australian English users are covered from the start, with more languages to follow.
One nice local detail: Voice Control rebuilt with Apple Intelligence is launching in English in the US, Canada, the UK and Australia, so we're on the first wave rather than waiting our usual turn. iOS 27 itself runs all the way back to the iPhone 11, which Apple reckons makes it the most widely supported iOS release ever. The Apple Intelligence and Siri AI features need newer silicon though: iPhone 16 or later (plus the iPhone 15 Pro and Pro Max), an M1 Mac or iPad or newer, or the latest Apple Watch and Vision Pro.
The verdict
For years Siri was the punchline in Apple's own keynote. Siri AI is the serious attempt to fix that, and on features alone it's comfortably the most capable assistant Apple has ever shipped.
The asterisk is who's underneath. Apple's whole brand is "your data never leaves your control," and it's now leaning on a model co-built with Google to keep that promise. Private Cloud Compute is a genuinely strong answer on paper, and Apple deserves credit for engineering a privacy story around a rival's AI rather than against it. Whether buyers care about the plumbing, or just want a Siri that finally understands them, we'll find out when this lands in spring.
My read: Apple has bought its way back into the AI race with Google's models and its own privacy moat. Pragmatic, a little awkward, and probably the right call. Now it has to ship something that lives up to the demo, and this time, actually keep it switched on for everyone.
Here is the keynote again:
Sources and further reading
Apple Newsroom, WWDC26 releases (all 8 June 2026):
Apple unveils next generation of Apple Intelligence, Siri AI, and more
Apple Intelligence brings powerful AI capabilities into everyday experiences
Apple introduces Siri AI, a profoundly more capable and personal assistant
Apple previews new child safety features
Apple aids app development with new intelligence frameworks and advanced tools
Due to DMA, Siri AI delayed in EU for iOS 27 and iPadOS 27
Apple expands App Store capabilities to help developers grow and reach new users
Background and context:
CNBC: Apple picks Google's Gemini to run AI-powered Siri
TechCrunch: Google's Gemini to power Apple's AI features like Siri
Google: Joint statement from Google and Apple
Digital Watch: Apple sues European Commission over DMA interoperability ruling
Free Software Foundation Europe: Apple keeps challenging its interoperability obligations under the DMA
Tom's Guide: Gemini vs Alexa+ tested
People:
Craig Federighi, Apple senior vice president of software engineering: Apple Leadership
Susan Prescott, Apple vice president of worldwide developer relations: LinkedIn
Dr Sumbul Desai, Apple vice president of health and fitness: Stanford Center for Digital Health
7 June 2026
4 June 2026
4 June 2026
3 June 2026
Leave a Reply