TL;DR
Apple unveiled Siri AI at WWDC 2026, a Gemini-powered rebuild with a standalone app, personal context search, and privacy-first cloud architecture.
Apple used its annual developer conference on Monday to unveil Siri AI, the most significant overhaul of its voice assistant in 15 years, rebuilt from the ground up on a custom Google Gemini model. The WWDC 2026 keynote at Apple Park also marked Tim Cook’s final appearance as CEO before he hands the role to hardware chief John Ternus on 1 September.
The rebuilt assistant arrives as a standalone app on iPhone, iPad, and Mac, functioning as a conversational chatbot alongside its existing system-wide presence. It can draw on personal context to search across messages, emails, and photos, execute multi-step commands across apps, answer questions about what is on screen, and go out to the web for up-to-date information. A dedicated Siri app uses iCloud to privately sync conversation history across devices.
Apple’s newsroom announcement framed the technology as “the next generation of Apple Intelligence” without naming Google, but multiple reports confirmed that Siri AI runs on a custom Gemini model of approximately 1.2 trillion parameters, licensed in a deal reported at roughly $1 billion per year. The architecture uses three tiers: simple tasks stay on-device using Apple’s own models, moderately complex requests run through Apple’s Private Cloud Compute, and the heaviest reasoning is routed to Google Cloud. Apple said that queries are processed statelessly, nothing is retained, and the contract bars Google from training future models on Apple user data.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
“We believe privacy in AI is non-negotiable,” Craig Federighi, Apple’s senior vice president of Software Engineering, said during the keynote. “Data is only used to execute your request, and outside experts can continue to verify this promise at any time.” Cook, in what multiple outlets described as a farewell tour, acknowledged that Apple Intelligence had “not yet delivered on everything we promised,” a rare public admission from a company that tightly controls its narrative.
That admission carried particular weight. Apple settled a $250 million consumer class action last month over marketing Siri AI features in 2024 that were not ready when the iPhone 16 launched. The personalised Siri capabilities originally advertised at WWDC 2024 were delayed indefinitely in March 2025. Monday’s keynote was, in practical terms, the delivery of features Apple had been sued for failing to ship.
Beyond Siri, the next generation of Apple Intelligence powers features across the system. In Photos, Spatial Reframing lets users improve composition after a shot is taken. Image Playground now supports photorealistic image generation. In Messages, one-tap suggestions can create notes or reminders from conversations. iOS 27 Extensions will let users set a third-party AI model, such as Claude or ChatGPT, as their default assistant, a concession to both competitive pressure and the EU’s Digital Markets Act requirements that previously forced Apple to pause its AI rollout in Europe.
Performance improvements across iOS 27 are substantial. Apps launch up to 30 per cent faster, photos load up to 70 per cent faster after being taken, and AirDrop transfers are up to 80 per cent faster. Browsing external drives on iPad is now up to five times faster, matching Finder on Mac. Apple also introduced a Liquid Glass slider in Settings that lets users adjust the transparency of the interface anywhere from ultra-clear to fully tinted, responding to criticism that last year’s glassy design was difficult to read.
The keynote devoted significant time to parental controls. A new child account setup immediately enables age-appropriate protections, with parents choosing which apps to make available. Contact approval requires parental sign-off for each new person a child connects with. Communication safety features automatically intervene if explicit or violent content is shared. A new Screen Time system lets parents set daily time allowances across Entertainment, Games, and Social Media categories, with default recommendations based on guidance from clinical and child development experts.
The software releases span iOS 27, iPadOS 27, macOS 27, watchOS 27, visionOS 27, and tvOS 27. Developer betas are available immediately, with public betas expected in mid-July and a general release this autumn alongside the next iPhone. On Apple Watch, a dynamic app grid surfaces five Siri-suggested apps and a new Find My app consolidates Find Devices, Find Items, and Find People. AirPods get custom EQ and expanded GymKit support. Apple Vision Pro users can turn panoramas into spatial environments, and Wi-Fi connection is up to three times faster.
The strategic question is whether Monday’s keynote resets the clock on Apple’s AI credibility or merely catches it up. Google, Meta, Microsoft, and a growing field of startups have been shipping conversational AI at scale for two years. Apple’s Gemini deal effectively acknowledges that it could not build a competitive large language model in-house on the required timeline. What it offers instead is distribution, with more than two billion active devices, and a privacy architecture that none of its competitors can match. Whether users notice the difference between a Gemini-powered Siri and ChatGPT or Claude will determine whether the $1 billion annual licence fee was worth it.
