Apple taps Google’s Gemini to rebuild Siri in 2026

Apple and Google have struck a multi-year AI partnership that will make Gemini the foundation for Apple’s next generation of “Apple Foundation Models,” a move expected to underpin a more personalized Siri in 2026. It’s a strategic trade: Apple accelerates capability by leaning on a best-in-class model stack, but takes on deeper dependency — and a new set of questions around control, data boundaries, and regulatory optics.

What happened

Apple and Google said on Jan. 12 that they have entered a multi-year collaboration in which the next generation of Apple Foundation Models will be based on Google’s Gemini models and cloud technology.

The companies said the models will support future Apple Intelligence features, including a more personalized Siri planned for 2026. Financial terms were not disclosed.

Key details

Apple continues to frame Apple Intelligence as a privacy-forward architecture: many requests are processed on-device, while more complex work can be routed to Private Cloud Compute (PCC), Apple’s server-side environment built on Apple silicon.

Apple says PCC is designed so user data is not retained and is not accessible to anyone other than the user — “not even to Apple.” Apple also says independent experts can inspect the code that runs on its Apple silicon servers, and that devices won’t connect unless the server software has been publicly logged for inspection.

This deal doesn’t eliminate Apple’s other model relationships. Apple has already integrated ChatGPT access across Siri and other experiences, with the system asking permission before sending prompts — and prompting again before photos or files are shared.

Why it matters

Apple’s Siri refresh is arriving under pressure. Apple has said some planned AI improvements to Siri slipped to 2026, and Siri handles roughly 1.5 billion user requests daily. At that scale, “revamping Siri” is not a feature swap — it’s a platform migration that has to work across languages, devices, and compliance regimes.

For Google, the benefit is distribution. Gemini becomes embedded in the assistant layer of Apple’s ecosystem, expanding reach across an installed base of more than two billion active Apple devices. The deal also builds on Gemini’s existing footprint in consumer devices, including deployments tied to Samsung’s Galaxy AI features.

For Apple, the win is speed and quality. The cost is structural: relying on an external foundation model raises the stakes on governance — who controls the model, where data flows, and how defaults are presented to users. Those questions matter more in an environment where platform defaults are already under antitrust scrutiny.

Governance is the product

Apple has historically differentiated through vertical integration — custom silicon, a tightly controlled OS, and services that reinforce hardware. Choosing Gemini as a foundation layer is a pragmatic pivot: buy frontier capability now, then wrap it in Apple’s privacy and UX scaffolding.

Whether that strategy succeeds comes down to governance in three places:

1. Model control: “Based on Gemini” leaves open who owns fine-tuning, safety policies, evaluation gates, and the cadence of updates. If model behavior shifts, the key question is whether Apple has hard controls — not just contractual promises — to maintain consistent Siri behavior and safety standards.
2. Data boundaries: Apple’s promise is that Apple Intelligence can be personal without exposing user data, thanks to on-device processing and PCC. As Gemini becomes foundational, Apple will need to show (in technical terms) what Google can’t see, what can’t be retained, and how those guarantees are enforced end-to-end. The joint statement does not spell out the operational boundary lines.
3. Regulatory optics: Defaults and bundling are no longer just a “search box” story. Google’s default-distribution strategy has already faced major antitrust rulings in U.S. search, and analyst estimates put Google’s annual payments to Apple for default search placement around $20 billion under an agreement that runs through at least September 2026. If regulators start treating the AI assistant’s intelligence layer as the next default choke point, an Apple+Google pairing could draw sharper scrutiny — even if the technical implementation is privacy-preserving.

Conclusion

Apple’s Gemini deal is a high-leverage move: it can raise Siri’s ceiling quickly, but it ties a core Apple experience to an external foundation model. The competitive battleground in 2026 won’t be model demos. It will be proof — proof of control over behavior, proof of hard data boundaries, and proof that users (and regulators) can understand and trust how the assistant is being powered.