Apple dumps in-house Siri AI, pays Google for Gemini
Apple has abandoned efforts to build proprietary artificial intelligence technology for its voice assistant and will instead pay Google to develop a customized system based on the Gemini platform. The arrangement will support computational tasks handled through Private Cloud Compute infrastructure when on-device processing proves insufficient.
Bloomberg reporter Mark Gurman disclosed the partnership shift in a recent newsletter. The updated assistant will feature three main elements: a query planner that routes requests efficiently, a knowledge database for general information queries, and summarization capabilities for text and audio content. Simple operations will run locally on user devices, while complex requests get transferred to Apple servers using encrypted connections, where the Google-built model will handle processing.
Engineers faced difficulties ensuring consistent performance across applications and sensitive scenarios like banking operations. The spring update for mobile software version 26.4 should introduce contextual app commands, personalized data integration and screen content recognition. Apple will present this technology as its signature product despite relying on external development partners. Additional enhancements are expected when version 27 arrives following the developer conference in June 2026.
