That posture later became concrete in coverage of [[a:1161418|Gemini-based Apple Foundation Models supporting Siri and future Apple Intelligence features]], making this interview an early marker of Apple’s model-partnership strategy.
First-order effects
The immediate effect is strategic signaling, not a product launch: Apple publicly leaves room for Gemini and other outside models inside its AI stack.
Google gains validation as a potential Apple AI partner, while Apple preserves flexibility to choose models by capability or use case.
Second-order effects
Other model providers face a clearer incentive to pursue Apple-compatible integrations, because the assistant interface can remain Apple’s even when underlying models vary.
Apple can compare external models against its own models without making the user experience synonymous with any one vendor; later reporting that Apple can fine-tune Gemini for Siri illustrates that division of roles.
Third-order effects
If this pattern persists, consumer AI platforms may increasingly separate the owning interface from the model supplier, shifting competition toward model performance, deployment terms, and integration quality.
The trade-off is greater supplier leverage for leading model vendors alongside Apple’s effort to retain control of the customer relationship; the eventual balance depends on how broadly Apple opens those integrations.
The trend: This is an early data point in the shift toward platform owners orchestrating multiple AI models behind a single branded assistant experience.
I'm *still* at WWDC and during an on-stage chat just now Apple's SVP of software engineering, Craig Federighi, said the company plans to integrate with AI models such as Google's Gemini “in the future”, and that different people will be able to use different models. interesting.
And of course, keep in mind that super-spies aren't your biggest adversary. For many people your biggest adversary is the company who sold you your device/software. This PCC system represents a real commitment by Apple not to “peek” at your data. That's a big deal. 20/
This sums up the difference between Apple and OpenAI's models. Apple's own model that runs on a device or its private cloud has a 3B parameter SLM. As a point of reference, GPT-4 LLM is estimated to be around a trillion. Bottom line: Apple, Microsoft and Amazon most advanced
Craig Federeghi says part one key for Apple is using generative AI for what it is ready to do today. “We're not taking this teenager and sort of telling him to go fly an airplane,” Federighi said.
Not clear yet if Apple's own AI models will clear Chinese government approvals, but at least the idea of letting users choose external models keeps the door open for third parties that have been approved in China #WWDC https://9to5mac.com/...
Giannandrea says Apple's llm was built in part using data from the public web and that publishers can opt out of, along with a wide range of licensed data. He doesn't get more specific, though.