Apple says its AI is still private, even when it’s running on Google’s servers

In a significant announcement from Cupertino, California, Apple has revealed that its much-anticipated Siri upgrade, now dubbed "Siri AI," will utilize Google’s Gemini language models. This development, highlighted during the recent Worldwide Developers Conference, marks a shift as the service will also leverage Nvidia hardware hosted on Google’s servers. Despite this transition, Apple is steadfast in its commitment to user privacy. The company reiterated its longstanding promise that privacy remains a top priority, even when the AI models operate on external servers. For years, Apple has championed its platforms as bastions of user privacy, utilizing encryption in its cloud services to safeguard data from unauthorized access, including from Apple employees. Apple has typically favored on-device processing to minimize data transmission, ensuring sensitive information stays within users' devices. However, as the demands for more robust language and reasoning capabilities have grown, Apple confronted the limitations of its own hardware. The language models that can run locally on devices like iPhones or Macs are relatively small, which constrains their performance and accuracy. To address this challenge, Apple has developed its Private Cloud Compute system, although it originally depended on Apple’s own server infrastructure. To effectively support Siri AI, Apple would have needed to invest heavily in expanding its data center facilities, a move the company has been cautious about. Following the keynote at WWDC, Apple executives, including Craig Federighi, took to a smaller stage to elaborate on the strategies they have in place to maintain user privacy while also meeting the computational demands of Siri AI. Their collaboration with Google aims to strike a balance between enhanced capabilities and the protection of user data.