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

Apple confirmed at its Worldwide Developers Conference that its upcoming Siri AI upgrade will utilize Google's Gemini language models and run on Nvidia hardware hosted on Google servers. This marks a significant shift from Apple's long-standing emphasis on user privacy, which has historically involved either on-device processing or computation on Apple-controlled servers. The company has consistently promoted privacy as a core benefit, employing encryption for its cloud services and prioritizing local data processing to minimize data leaving user devices. However, the computational demands of advanced AI models like those powering Siri AI have outpaced the capabilities of current on-device hardware. Apple's previous solution, Private Cloud Compute, relied on Apple's own server infrastructure. To meet the required capacity for Siri AI, Apple would have needed to invest heavily in data center expansion, a move it has thus far avoided. The integration with Google's infrastructure allows Apple to leverage substantial computing power without undertaking a massive hardware buildout, while still aiming to uphold its privacy commitments through its Private Cloud Compute system.