iOS 27 Will Let Users Swap Apple Intelligence for Rival AI Models

Image: Bloomberg AI
Main Takeaway
Apple plans to open Siri and system-level AI features to third-party models like ChatGPT, Gemini, and Claude in iOS 27, marking a dramatic shift from.
Jump to Key PointsSummary
A radical reversal of Apple's AI strategy
Apple will abandon its one-size-fits-all approach to AI this fall when iOS 27, iPadOS 27, and macOS 27 ship with a new “AI preference” pane that lets users swap Apple Intelligence for any supported third-party chatbot, according to Bloomberg managing editor Mark Gurman. The move ends months of speculation that Apple would lock its devices to a single provider after it struck a ChatGPT partnership last year. Instead, users will be able to designate ChatGPT, Google Gemini, Anthropic Claude, or other future models as the default engine for Siri, writing tools, photo queries, and any other Apple Intelligence surface.
What this means for developers
App makers gain a single API to reach whichever model the user has chosen, eliminating the need to embed multiple AI SDKs. Apple will reportedly expose a lightweight abstraction layer that normalizes prompts and responses across providers, so a note-taking app can ask for “summary” without caring whether the underlying model is GPT-5 or Gemini 2.0. Early documentation seen by Bloomberg suggests Apple will not charge a commission on AI calls made through this layer, a departure from its usual 15-30 percent take on digital goods.
Why this matters for open source
By refusing to privilege its own models, Apple effectively turns every iPhone into a neutral playground for competing AIs. Open-source teams can ship lightweight “AI drivers” to the App Store that plug into the same slot as commercial giants, giving users a path to run Llama-4 or Mistral on-device without jailbreaking. The policy also pressures closed providers to publish clear privacy and pricing terms, since users can switch with two taps.
The impact on enterprise adoption
Corporate IT departments have long resisted Apple Intelligence over data-residency fears. Letting them choose, say, an on-prem Claude model or a Microsoft Azure-hosted GPT removes that blocker. Sources tell Bloomberg that Apple will ship an MDM payload to pre-select an enterprise AI for supervised devices, making fleet rollouts trivial. The change could accelerate iPhone refresh cycles in regulated industries that have so far stuck with Android for AI flexibility.
What happens next
Gurman expects Apple to detail the AI preference system at WWDC 2026 on June 8, alongside a redesigned Siri interface and new Visual Intelligence camera tricks. Third-party model makers must pass Apple’s privacy audit and agree to on-device processing for sensitive requests, but no exclusivity deals are required. The first wave of supported partners will reportedly include OpenAI, Google, Anthropic, and Perplexity, with more added quarterly. Apple Intelligence itself will remain an option, just not the only one.
Key Points
iOS 27 will add an "AI preference" pane to choose any supported third-party chatbot as the system-wide assistant.
Developers gain a single, commission-free API that abstracts away which model the user has selected.
Enterprise IT can pre-select an on-prem or cloud model via MDM, easing regulated-industry rollouts.
Open-source models can ship lightweight drivers to the App Store and plug into the same slot as commercial AIs.
First partners to be announced at WWDC 2026: OpenAI, Google, Anthropic, and Perplexity.
Questions Answered
Apple requires all models to process sensitive requests on-device; cloud queries must pass a privacy audit, but the partner—not Apple—handles the data.
The initial implementation supports one global default, though Apple may expand to task-specific choices in later updates.
No; Apple provides a single abstraction layer so the same code works regardless of which model the user has chosen.
Apple charges no commission, but individual providers may apply their own subscription or per-token pricing.
Source Reliability
36% of sources are trusted · Avg reliability: 69
Go deeper with Organic Intel
Simple AI systems for your life, work, and business. Each one includes copyable prompts, guides, and downloadable resources.
Explore Systems