Apple ships Gemini-powered Siri beta with iOS 27, excluding EU and China at launch

Apple has shipped a beta of a rebuilt Siri, powered by Google Gemini models running in a hybrid on-device and Private Cloud Compute setup, as part of the iOS 27/iPadOS 27/macOS 27/watchOS 27/visionOS 27 release line. New capabilities include reading on-screen content, drawing personal context from messages and photos, cross-app actions, third-party integrations (WhatsApp and Audible now, Outlook, Notability and Tripsy to follow), new voices and a new dictation model (AFM Core Advanced). It is English-only at launch, with French, Japanese, Korean, Portuguese and Spanish due next month, and the EU and China are excluded from the rollout entirely.

This is a real, shipped dependency: Apple has put a Google model at the centre of the assistant layer on roughly a billion devices, which matters for anyone tracking the balance of power between the two companies and for developers deciding where to integrate. The EU exclusion is a concrete regulatory data point worth tracking, likely tied to DMA compliance costs or interoperability requirements rather than a technical limitation. Real-world impact for ordinary users is still unproven, since this is a beta with reported accuracy and personal-context problems, not a finished product.

The "Apple finally closed the AI gap" framing is Apple's own launch narrative being echoed rather than tested; three hands-on reviews cited report failures alongside wins, so the gap is narrowed, not closed. The underlying source here is secondary, a write-up of Apple's announcement, not primary documentation, so financial or technical details of the Apple-Google arrangement are absent and should not be assumed.

1. 01Whether Apple discloses commercial terms of the Gemini deal (licensing fee, revenue share, or per-query cost) in an earnings call or SEC filing
2. 02Whether the EU version ships within the following two quarters and what changes (or does not) versus the global version, as a signal of what the exclusion was actually about
3. 03User-facing accuracy and personal-context complaint volume once the beta reaches general release, via App Store reviews or independent benchmarks rather than early hands-on pieces