Apple introduced Siri AI at its WWDC 2026 keynote on Monday in Cupertino, a ground-up rebuild of the assistant that now runs on Google's Gemini models. It ships this fall with iOS 27 and five other operating systems, and it was the centerpiece of Tim Cook's final keynote as chief executive.
The framing Apple chose tells you how the last two years went. Cook walked on stage to sell a Siri that, by the company's own admission at WWDC 2024, was supposed to already exist. That demo slipped. Then it slipped again. The fix, it turns out, was paying someone else to build the brain.
So Apple gave up and called Google
The partnership is the actual story here, and Apple is being unusually direct about it. The company said it worked with Google and the Gemini family of models to build the next generation of Apple Foundation Models. Reports put the deal at roughly a billion dollars a year, though Apple hasn't confirmed a figure and you should treat that number as a leak, not a line item.
Craig Federighi, Apple's software chief, got ahead of the obvious criticism by arguing the new Siri is more deliberate than rivals who, in his telling, bolt on AI just to say they have it. Maybe. It's a strange flex to deliver while announcing your assistant is powered by a competitor's foundation model.
Tasks now route across three tiers: simple stuff stays on the device, heavier work goes through Apple's Private Cloud Compute, and the most demanding queries reach Gemini. Apple's pitch is the same one it always makes, that your data is used only for the request at hand and stays out of reach of Apple and third parties. The privacy claim is plausible and also unverifiable from the outside, which is roughly where Apple wants it.
What you actually get
The voice is the easy win. Siri AI is more conversational, with adjustable pacing and expressiveness, and it sounds less like a synthesizer reading a phone tree. There's a standalone chatbot app now, the thing Bloomberg had been reporting for months, where you can ask questions, generate text and images, and hand it files to analyze. Conversations sync across iPhone, iPad, and Mac.
Visual intelligence lets Siri work with what the camera sees or what's on your screen. Point it at a restaurant bill to split the check, or a meal to pull nutrition info. On the Mac, it lives inside Spotlight, and one demo had it comparing a stack of files and spitting out a chart, then drafting an email to place an order. Useful if it works as cleanly outside a scripted keynote, which is a big if.
The fine print on who gets it
Here's where the rollout gets messy. Siri AI runs on iPhone 16 and later, plus the iPhone 15 Pro and Pro Max, iPad mini with A17 Pro, the new MacBook Neo, and any iPad or Mac on M1 or later. On the watch side it's Series 9 and up, Ultra 2 and up, and the SE 3, all tethered to a nearby compatible iPhone. The fastest experience, with the expressive voices and better dictation, wants an iPhone 17 Pro, 17 Pro Max, or iPhone Air.
Two big regions wait. Siri AI won't ship in the EU on iOS 27 or iPadOS 27 at launch, which Apple blames squarely on the Digital Markets Act, saying regulators rejected its proposals for supporting third-party assistants safely. EU users on Mac, Watch, and Vision Pro can still get it in a supported language. China is out entirely at launch while Apple works through local rules.
It launches in English only. Developer betas are out now, public betas land in July, and the full release arrives in the fall. Cook hands the CEO job to hardware engineering chief John Ternus on September 1, which means the man who announced Siri AI won't be running the company when it ships.
