APREX · EST. 2026
Apple has spent years building its brand on privacy. Now the company is challenging its own narrative by tying Siri more closely to one of the world's largest data collectors: Google.
A Billion-Dollar Partnership With Big Ambitions
At WWDC 2026, Apple unveiled its forthcoming overhaul of Siri, with Google's Gemini technology serving as the foundation for the most demanding requests. According to Wired, the partnership is estimated at around one billion dollars and will span several years.
Structurally, the collaboration resembles the existing ChatGPT integration: the Apple device handles simple tasks locally, while more complex queries are routed to Google's infrastructure via Apple's own cloud services — not directly to Google.
How the Architecture Works
Apple emphasizes that Gemini does not run directly on the user's iPhone. Instead, Apple's own models handle on-device processing, while complex requests are sent through Apple's Private Cloud Compute infrastructure to Gemini. Craig Federighi, Apple's software chief, claims the company uses models that are not the same ones Google rolls out to its other customers, and that user requests are "never stored and never accessible to others."
Apple also ties Siri queries to a random, device-based identifier — not the user's Apple account or phone number.
Users are trusting both companies to honor their commitments — and that is no small thing to ask.
EU and China Left Out
The new Siri features will not be available in the EU, where GDPR imposes stricter requirements on data processing, nor in China. Apple has not yet communicated any concrete timeline for a potential expansion to these markets.
Experts Are Not Convinced
Despite assurances from both companies, privacy experts remain skeptical. Jennifer King at the Stanford Institute for Human-Centered AI notes, in comments to research sources, that sensitive information shared with Gemini models may be collected and used for training even when users believe they are in control. She generally advises against sharing confidential information with AI chatbots.
Google's own guidelines for the Gemini service indicate that conversations may by default be used for model training, with an opt-out option available. Conversations can also be retained for up to three years if they have been reviewed by human reviewers — and this applies even after the user has deleted them.
Apple's Privacy Track Record Under Scrutiny
This is not the first time Apple has faced criticism over Siri-related privacy. In 2025, the company reached a 95 million dollar settlement following allegations that Siri had recorded conversations without explicit user consent — an issue that first came to light in 2019 when it emerged that contractors had been listening to recordings. Apple suspended human review and introduced mandatory consent as a result.
Taken together with the new Google deal, this paints a picture of AI service privacy being far more complex than companies' own communications departments tend to let on. Users would do well to read the terms carefully — and to bear in mind that technology partners do not always share the same incentives.
Sources: Wired, Stanford HAI (via research)