At WWDC and on its product page, Apple unveiled Siri AI, a rebuilt assistant with on-screen awareness, broader access to on-device context, and the ability to take actions across apps; Apple presents these features as part of Apple Intelligence, with new capabilities arriving "coming this fall" and "Siri AI coming in English later this year" (Apple). Reporting by multiple outlets shows hands-on testers were able to ask Siri to extract calendar events from email, edit text across apps, and add reminders (The Verge; Wirecutter). Independent reporting claims Apple rebuilt Siri on Google's Gemini, reportedly a custom 1.2-trillion-parameter model and that Apple pays Google on the order of $1 billion per year for access (Cringely; Wired reporting). Editorial analysis: granting an assistant the ability to act inside apps increases operational risk from hallucinations and bad actions, even if privacy protections are emphasized.
What happened
Apple revealed a rebuilt assistant, branded Siri AI, at WWDC and on its product page for Apple Intelligence. Per Apple's product page and WWDC demos, Siri AI will have on-screen visual awareness, access to on-device context across Photos, Messages, Mail, Calendar, and the ability to take actions inside third-party and first-party apps such as editing messages, adding calendar events, and managing Safari tabs (Apple; Wirecutter). Apple lists the release timeframe as "coming this fall" and "Siri AI coming in English later this year" (Apple). The Verge and other hands-on reporters verified several of these features in early builds, including extracting events from email and composing or editing across apps (The Verge; MacRumors; TechRadar).
Technical details
Reporting from Cringely and coverage in Wired state that Apple's new assistant is rebuilt on Google's Gemini, with Cringely claiming a reportedly custom 1.2-trillion-parameter variant and an annual payment to Google on the order of $1 billion. These vendor-relationship and model-size claims are reported by those outlets and are not stated on Apple's public product page; Apple's marketing materials instead emphasize a combination of on-device context plus broader world knowledge via cloud services (Cringely; Wired; Apple).
Industry context
Editorial analysis: Companies that enable assistants to act inside user workflows enlarge the attack surface for operational errors. In previous, chat-only deployments a hallucination typically costs time and a correction. When an assistant can perform transactions, edit content, or move items between apps, hallucinations can produce tangible errors: mis-sent messages, incorrect calendar entries, or undesired file deletions. Industry reporting frames Apple's pitch as a privacy-forward counterpoint to competitors, but independent coverage highlights that privacy assurances and action-capability create different risk profiles that product teams must manage (Fortune; Cringely).
What to watch
Editorial analysis: Observers should track three indicators: adoption and coverage of the Gemini integration in independent tests; how Apple constrains or logs agent actions across apps (platform-level APIs, permission prompts, undo/consent UX); and the frequency and severity of action-driven errors reported by early users and reviewers. Also watch enterprise guidance and documentation from Apple about developer controls for Siri AI in third-party apps.
Quoted context
Fortune reproduces Apple software chief Craig Federighi's WWDC remark criticizing a race-for-AI approach: "Some appear to be racing forward, seemingly pursuing AI for the sake of AI, without clear regard for the people, all of us, that it's ultimately meant to serve" (Fortune). That quote frames Apple's public positioning but does not specify technical guardrails for action-taking assistants.
For practitioners: integrating an assistant with hands emphasizes robust end-to-end testing, more conservative defaults for action permissions, and clearer audit trails. Industry teams deploying assistant-driven automation should expect to instrument action confirmations, rollback paths, and monitoring that treats generated actions as first-class events for reliability and safety analysis.
Scoring Rationale #
A major consumer-platform vendor shipped an assistant that couples rich on-device context with cross-app actions, which materially raises operational and design requirements for practitioners. The story is important but not a frontier-model paradigm shift.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.