Apple plans to make iOS 27 a Choose Your Own Adventure of AI models

The framing of user choice obscures the more consequential shift: Apple would become an inference broker, sitting between users and model providers, which gives it distribution leverage without the liability of owning the model outputs. That intermediary position is the real strategic asset here, not the optionality itself.

This connects directly to the Pentagon's multi-vendor AI procurement reported in early May, where institutional concern about single-vendor dependency drove explicit redundancy into the stack. Apple is applying the same logic at the consumer OS layer. It also rhymes with the enterprise 'AI factories' framing from MIT Technology Review's EmTech coverage, where decentralized model control was positioned as a sovereignty hedge. Meanwhile, OpenAI's move to ad-supported free tiers (covered here from The Decoder, May 2) makes model interchangeability more commercially urgent for Apple: if OpenAI's free tier becomes surveillance infrastructure, Apple has brand reasons to offer alternatives, not just competitive ones.

Watch whether Apple publishes a formal inference API specification before iOS 27 ships. If it does, that confirms the broker-layer thesis and forces Google and OpenAI to decide whether to certify against Apple's standard or risk exclusion from the default chooser.