Apple completely rebuilds Siri: Google's Gemini becomes the brain of 1.4 billion iPhones - users will be able to choose between three AI models
What it really says
At WWDC 2026 on June 8, Apple unveiled the most comprehensive overhaul of Siri since the voice assistant's introduction in 2011. At its core is a licensed Google Gemini model with 1.2 trillion parameters in a mixture-of-experts architecture that activates only a relevant subset of parameters per query. The model is approximately eight times larger than the largest cloud model Apple has built on its own. According to Bloomberg, Apple pays Google roughly one billion US dollars per year for it. The new Siri understands context across multiple conversation turns, can execute multi-step tasks across different apps, and has access to personal data such as emails, photos, and files. Activation occurs via a new swipe-down gesture and Dynamic Island integration. Particularly noteworthy is the new Extensions system: users will be able to choose which AI model powers their Apple Intelligence features - alongside Gemini, OpenAI's ChatGPT and Anthropic's Claude are also available. This effectively turns the iPhone into an AI distribution center where the user retains control over the provider. The Gemini model runs within Apple's Private Cloud Compute infrastructure on Apple Silicon servers with stateless, ephemeral processing: no user data is retained after a query, and Apple's contract explicitly prohibits Google from using Apple user queries to train future Gemini models. Additionally, iOS 27 brings three generative tools to the Photos app, all running on-device: Extend for generative fill, Enhance for smart auto-adjust, and Reframe for shifting perspective on spatial photos. This is Tim Cook's final WWDC keynote as CEO.
Our assessment
For people concerned about AI in everyday life, this announcement is a double-edged sword. On one hand, Apple is making AI the default for 1.4 billion iPhone users - anyone using an iPhone will hardly be able to avoid AI interaction going forward. The deep integration into emails, photos, and personal files means an AI model potentially gains access to the most intimate digital areas of one's life. On the other hand, Apple deserves credit for several design decisions: the Private Cloud Compute architecture with stateless processing is an industry-leading privacy approach. Google receiving no training data from Apple queries is contractually secured. And the choice between three AI providers gives users a level of control that exists with no other smartphone manufacturer. The critical question remains: how transparent is the data flow between Apple's servers and the Gemini model really? Independent audits of the Private Cloud Compute infrastructure are still pending. Moreover, the dependence on Google represents a strategic shift: Apple, which positioned itself as a privacy champion for years, is now handing over the core intellectual capability of its assistant to the world's largest data collector. The price tag of one billion dollars per year shows that Apple is unable to run a comparable model on its own for the foreseeable future. For the average user, the practical benefits outweigh the risks: Siri is finally becoming useful. The privacy risks are real, but better mitigated by Apple's architecture than by most competitors.
Relevance for Germany
Germany is one of Apple's most important markets after the US and China, with an estimated 30 million active iPhone users. This means the Gemini-powered Siri will enter millions of German households and businesses. For German data protection, the Private Cloud Compute architecture is fundamentally positive, as it appears compatible with the GDPR's data minimization principle. However, data protection authorities like the BfDI will want to closely examine whether stateless processing is truly guaranteed and whether the servers are located in the EU or the US. The choice between Gemini, ChatGPT, and Claude is also relevant from a competition law perspective: it could serve as a model for the Digital Markets Act, which requires dominant platforms to offer users alternatives. For German workers, the Siri overhaul means that AI-powered assistants are now becoming standard in the Apple ecosystem as well - another step in the normalization of AI in the workplace. It is also noteworthy that the Extensions system could theoretically be open to German AI providers, which would represent an opportunity for companies like Aleph Alpha.
Fact check
The announcement comes directly from Apple's WWDC 2026 keynote on June 8, 2026. The technical details - 1.2-trillion-parameter Gemini model, mixture-of-experts architecture, Private Cloud Compute infrastructure with stateless processing - are consistently reported by MacRumors, Tom's Guide, TechTimes, Apfeltalk, and Enterprise DNA. The licensing cost of approximately one billion dollars per year comes from a Bloomberg report cited by multiple sources. The contractual assurance that Google cannot use Apple user data for training is confirmed by MacRumors and RedShark News. The Extensions system with the choice between Gemini, ChatGPT, and Claude is consistently described by MacRumors, TechCrunch, and Apfeltalk. The figure of 1.4 billion active iPhones comes from Apple's own investor reports.
Source
- • https://www.macrumors.com/guide/wwdc-2026-what-to-expect/
- • https://www.tomsguide.com/news/live/wwdc-2026-live-news-updates
- • https://www.techtimes.com/articles/317902/20260606/wwdc-2026-opens-monday-gemini-powers-rebuilt-siri-iphone-11-faces-ios-27-cut.htm
- • https://www.apfeltalk.de/magazin/news/wwdc-2026-was-ihr-von-siri-ueberarbeitung-und-apple-intelligence-neuerungen-erwarten-koennt/
- • https://enterprisedna.co/resources/news/apple-wwdc-2026-gemini-siri-ios27-enterprise-2026/