Apple Showcases Latest Siri AI Technology, Revealed to be Powered by Google
Apple officially showcased its latest artificial intelligence (AI) capabilities, including the newest version of the virtual assistant Siri, at the WWDC event held in Cupertino, California. However, behind this major announcement, it was revealed that the AI platform, dubbed ‘Apple Intelligence’, was not developed entirely in-house but through partnerships with Google and Nvidia to develop its core technology.
At WWDC, Apple demonstrated a transformed Siri. Unlike previous versions, Siri is now capable of seamless two-way conversations with users. In one use case, the assistant can check concert schedules, set reminders to purchase tickets, and find travel routes while simultaneously organising pick-up plans for friends en route to an event.
This move underscores Apple’s distinct strategy compared to many of its competitors. Rather than spending billions of dollars to build massive infrastructure and develop the most advanced AI models independently, Apple is instead emphasising advantages in privacy and user convenience.
“Some parties seem to be racing forward, as if chasing AI solely for the sake of technology itself, without deeply considering all of us—the society served by this technology,” said Apple’s SVP of Software, Craig Federighi, during his opening speech.
While claiming to have its own development path, Apple confirmed that its most advanced AI model, named Apple Foundation Model Cloud Pro (AFM Cloud Pro), was developed with the assistance of technology from Google and Nvidia.
The partnership with Google was actually announced last January, but this is the first time Apple has officially acknowledged that several flagship features of its AI system, known as Apple Intelligence, will run using Nvidia-made processors.
Apple AI executive, Amar Subramanya, explained that the capabilities of AFM Cloud Pro are comparable to Google’s advanced Gemini models. This model operates on Nvidia GPU-based cloud services, which are integrated into Apple’s proprietary data security infrastructure known as Private Cloud Compute.
“We are working with Google and Nvidia to extend the reach of our security infrastructure to existing Nvidia GPU devices within Google’s cloud services, while still guaranteeing the highest standards of data protection that are characteristic of Apple,” said Subramanya.
The primary differentiator highlighted by Apple is their approach to data. Unlike web-based AI services such as OpenAI’s ChatGPT or Anthropic’s Claude, which process vast amounts of data on external servers, Apple claims to collect less user data. Instead, the system utilises information stored locally on devices, such as calendar schedules or text messages, to provide a personalised yet secure experience.
Federighi also clarified the boundaries of the partnership with Google. He emphasised that Apple Intelligence does not use the same Gemini technology available to the general public, nor does it use Google’s standard cloud infrastructure.
Google’s role is primarily in the development and training stages of the models. “The third-generation models we are launching are specifically designed to run smoothly on devices with Apple-made processors. These models are trained using internal data and refined using learning results from Google’s advanced models,” explained Federighi.
Subramanya added that the four main models introduced—AFM Core, Core Advanced Cloud, and Cloud Image—are all custom-built for Apple devices, trained with specialised learning methods, and refined with the assistance of Google’s artificial intelligence technology.
In this manner, Apple seeks to balance the need for cutting-edge AI technology with its primary commitment to keeping users’ personal data secure, even when collaborating with other tech giants.