Apple readies Siri overhaul powered by Google Gemini

Apple unveiled Siri AI at WWDC 2026 on June 8, 2026, the biggest overhaul of the assistant since its 2011 debut, arriving as part of the next-generation Apple Intelligence across iOS 27, iPadOS 27, macOS 27, watchOS 27, visionOS 27, and tvOS 27 (Apple Newsroom). The rebuilt assistant holds multi-turn conversations, pulls context across apps like Mail, Messages, Photos, and Calendar, understands on-screen content, and can execute multi-step tasks autonomously. A new standalone Siri app lets users revisit past conversations, start new ones, and attach images or documents, with history synced privately across devices via iCloud; users can set conversations to auto-delete after 30 days or a year, or keep them indefinitely, mirroring Messages' retention controls (Apple Newsroom; TechCrunch). Siri AI ships as an English-only beta later in 2026, expanding to more languages afterward, with full release alongside new iPhone hardware this fall.

Apple says the new capabilities run on its third-generation Apple Foundation Models: two on-device models (AFM 3 Core and AFM 3 Core Advanced) and three server-based models (AFM 3 Cloud, AFM 3 Cloud Image, and AFM 3 Cloud Pro), with Cloud Pro handling the most demanding agentic and complex-reasoning tasks (Apple Machine Learning Research). To run Cloud Pro, Apple extended Private Cloud Compute onto Google Cloud infrastructure using NVIDIA GPUs with Confidential Computing, Intel CPUs with TDX, and Google's Titan chip, while keeping PCC's transparency and verification guarantees intact (security.apple.com). Apple's official materials describe this as building new Apple models using "the technologies behind" Google's Gemini family, and AppleInsider has disputed widely circulated claims that Siri simply "runs on Gemini," arguing the resulting models are Apple's own rather than a repackaged Gemini. Bloomberg's earlier reporting, cited widely including by CNBC, put the commercial arrangement at roughly $1 billion a year for a custom, approximately 1.2 trillion-parameter model; Apple has not confirmed either figure.

Apple confirmed Siri AI will not launch in the European Union or China alongside the rest of the iOS 27 rollout. Apple says EU regulators rejected its proposed approach for extending Siri AI's on-device access under the Digital Markets Act while safeguarding other assistants, and it has no timeline for bringing the iOS/iPadOS version to the EU, though EU users get it on macOS 27 and visionOS 27 (Apple Newsroom; MacRumors). In China, the delay reflects the separate requirement that generative AI models clear local regulatory approval before public release.

The AFM 3 lineup and its five-tier on-device/server split is a concrete reference architecture for teams weighing when to keep inference local versus route to a cloud-hosted model tier, and Apple's cross-cloud PCC extension is a live example of enforcing a security and transparency boundary on infrastructure a company doesn't fully control. Anyone building assistant features that pull cross-app personal context should also treat Apple's carve-outs for the EU and China as a preview of how autonomous, wide-access AI agents are likely to be scrutinized under the DMA and China's generative-AI rules.

• •Whether Apple and Google disclose confirmed commercial terms, closing the gap between Bloomberg's reported $1 billion/1.2 trillion-parameter figures and Apple's silence on both.
• •How the EU dispute resolves, and whether Apple proposes a new DMA-compliant design for Siri AI on iOS/iPadOS.
• •China's regulatory review timeline for AFM 3 Cloud Pro and whether Apple ships a China-specific model, as it has for prior on-device Apple Intelligence features.
• •Developer access: whether Apple's WWDC26 sessions expose new Siri App Intents or APIs beyond what's in the initial beta.