At WWDC 2026, Apple unveiled significant advancements in its Apple Intelligence platform, particularly highlighting the enhanced Siri AI application. Following the event, intriguing details have surfaced regarding the collaboration between Apple and Google, specifically how Google's Gemini AI models are being integrated into Apple's software. However, the distinction between what is uniquely Apple and what is derived from Google remains complex.
Earlier this year, Apple announced its multi-year partnership with Google to utilize Gemini AI models, ensuring that its renowned privacy standards would remain intact. This collaboration marks a pivotal shift from Apple's previous approach, where Siri would redirect prompts it could not address, to a more sophisticated AI integration.
Siri AI: Distinct from Gemini
During the keynote, Craig Federighi, Apple's Senior Vice President of Software Engineering, emphasized that the newly developed Apple Foundational Models (AFM) are the result of a "deep collaboration" with Google. He noted that Apple has been "leveraging" Gemini technology to enhance Siri and other features within Apple Intelligence.
The impact of this collaboration is evident: Siri AI has evolved into a multi-modal system, adept at processing audio, voice, and text, and it now excels in generating both text and images. The improvements in image editing are particularly noteworthy, showcasing capabilities akin to those of advanced models like Nano Banana 2. Additionally, Apple's AI now possesses enhanced world knowledge, an area where Gemini models excel.
Despite this partnership, it's crucial to note that these advancements are fundamentally Apple's own models. The on-device AFM 3 Core and AFM 3 Core Advanced models are designed to operate on iPhones, iPads, and Macs, requiring the latest hardware for optimal performance.
In follow-up remarks, Federighi clarified that Apple Intelligence does not utilize the Gemini app or its associated client code. The development and operational framework have been exclusively managed by Apple, reinforcing the company's commitment to maintaining its proprietary identity while benefiting from Google's technological prowess.
Utilizing Google's Infrastructure
While Apple maintains local models that prioritize user privacy, the cloud-based models introduce a layer of complexity. Apple's AFM 3 Cloud models are designed for general use, including image generation and more advanced queries, with some processes being handled on Google's data centers utilizing Nvidia GPUs.
Importantly, Apple assures that its Private Cloud Compute protections extend to data processed on Google servers, ensuring that no data is stored post-query and that user identities remain confidential. Apple is committed to transparency, allowing third-party audits of its security protocols.
As Apple continues to refine its AI capabilities, the integration of Gemini models appears to strike a balance between enhancing functionality and preserving the unique essence that users expect from the brand. This collaboration not only positions Apple to compete more effectively in the AI landscape but also sets the stage for future innovations in personal assistant technology.
