Apple is routing some of its AI model processing through Google's cloud infrastructure, and claiming Google still has no access to user data.
Apple Intelligence, the company's suite of on-device and server-side AI features, has always relied on what Apple calls Private Cloud Compute for tasks too demanding for a phone's chip. That architecture is designed so processing happens without Apple's own engineers being able to read the data in transit or at rest. Now some of that compute is landing on Google Cloud, and Apple says the same privacy guarantees carry over - Google's infrastructure handles the workload, but Google itself sees nothing.
This is a stress test for one of Apple's most valuable brand assets. Apple has spent years marketing itself as the company that doesn't mine your data, a pitch aimed squarely at ad-funded rivals - Google chief among them. Offloading AI jobs to Google's servers, even with claimed technical isolation, asks users to trust a privacy promise brokered through one of their main supposed antagonists. If the architecture holds up to independent scrutiny, it could become a legitimate template for privacy-preserving cloud outsourcing.
Apple made a similar argument when it previously integrated OpenAI's ChatGPT into Siri, but OpenAI isn't in the business of building advertising profiles. Google is. That's a meaningfully different kind of trust exercise, and Apple's say-so alone probably shouldn't be enough to close it.