Apple could let you pick a favorite AI model in iOS 27

The buried angle here is distribution economics, not user choice. If Apple routes system-wide queries through third-party models, it becomes a gatekeeper with the power to set terms, take a cut, or preference certain providers, which is a different kind of lock-in than the one it's ostensibly dismantling.

This connects directly to the ethical divergence story from The Decoder (early May), which found that different frontier models encode meaningfully different value systems. If iOS 27 lets users swap models system-wide, Apple inherits a governance problem: whose ethics does Siri enforce when the underlying model is user-selected? That story showed enterprises already face implicit moral choices when picking models, and Apple would be pushing that ambiguity down to individual consumers. Separately, the OpenAI ad-tracking shift from May 2 is relevant context: a free-tier ChatGPT that monetizes through behavioral tracking looks considerably less attractive as a default iOS option, which could actually benefit paid-tier providers and reshape which models gain traction through this channel.

Watch whether Apple announces any certification or content-policy requirements for third-party models at WWDC 2026. If it does, that confirms the gatekeeper dynamic is real and will determine whether this is genuine optionality or a curated marketplace with Apple controlling the velvet rope.