Apple is planning to let users choose which AI model powers Apple Intelligence in iOS 27, iPadOS 27, and macOS 27, all expected to ship in fall 2026. The mechanism is a new class of plug-ins called Extensions, which would let third-party chatbots run Siri, Writing Tools, and Image Playground system-wide. Bloomberg's Mark Gurman reported the plan on May 5, 2026, and said Apple is already internally testing integrations with Google and Anthropic models.
The structural change matters more than the feature list. Today, ChatGPT is the only outside model wired into Apple Intelligence, and only as a handoff for queries Siri can't answer. Extensions would push that integration deeper — into the writing and image features Apple ships across the OS — and break ChatGPT's exclusive status in the process.
According to Gurman, users will be able to choose from AI model providers that opt in by adding support through their App Store apps. Once a compatible AI app is installed, the user picks it as their preferred AI through the Settings app, the same flow Apple uses for default browsers and mail clients. It's a model-agnostic posture Apple has historically resisted at the OS level.
Key facts
- 01Apple plans to let third-party chatbots power Apple Intelligence system-wide in iOS 27, iPadOS 27, and macOS 27, all expected fall 2026.
- 02Extensions will run Siri, Writing Tools, and Image Playground — not just the ChatGPT handoff that exists today.
- 03Users will pick a default AI in Settings after installing an opted-in App Store app, and can assign different Siri voices per model.
- 04Mark Gurman reports Apple is already internally testing integrations with Google and Anthropic models.
- 05Google's models are also slated to power Apple's upcoming Siri revamp, ending ChatGPT's exclusive status.
Apple is also planning to let users assign different Siri voices to different AI models. A Siri response from one of Apple's in-house models could speak in one voice; an instance of Siri running on ChatGPT could use another. The detail sounds cosmetic but solves a real product problem — telling the user which brain just answered them.
“Compatible third-party AI models, called Extensions, will be able to run Siri, Writing Tools, and Image Playground — features locked to ChatGPT and Apple's own models today.”— Jaeden Schafer
The list of internal test partners is the most telling part of the report. Google and Anthropic are the two names Gurman cites, and Google's models are also the backbone of Apple's upcoming Siri revamp. Apple is hedging on the model layer the same way it hedges on chip suppliers: pick more than one, keep them competing, and don't let any single vendor define the user experience.
For Anthropic and Google, distribution through Apple Intelligence is the prize. Apple's installed base is the largest premium consumer surface in software, and an Extension slot puts Claude or Gemini one Settings toggle away from being the default assistant on a billion-plus devices. That's a different acquisition channel than chatbot apps competing for downloads.
For OpenAI, the change cuts the other way. The exclusive ChatGPT handoff Apple announced in 2024 has been one of the highest-profile distribution wins in the AI industry, and Extensions effectively ends the exclusivity. ChatGPT becomes one option in a list rather than the only outside brain Apple endorses.
There are caveats. Gurman's reporting describes internal plans, not a shipped product, and Apple has pulled features from iOS releases before — the personalized Siri overhaul slipped from iOS 18 into a multi-year project. Extensions could narrow in scope, slip past fall, or launch in some markets before others. The opt-in requirement also means the lineup at launch depends on which providers actually ship App Store apps wired to the new APIs.
Privacy is the other open question. Apple Intelligence's pitch has rested on on-device processing and Private Cloud Compute, and routing Siri queries to Google or Anthropic's servers complicates that story. Apple will need to be explicit about what leaves the device when a user picks a third-party Extension, and the disclosure design will shape how many users actually flip the switch.
The strategic read is that Apple has decided it doesn't need to win the model layer to win the AI layer of its platform. By turning model choice into a Settings menu, Apple keeps the interface, the distribution, and the trust relationship with the user, while letting OpenAI, Google, and Anthropic fight for the slot underneath. It's the same playbook Apple ran on search, where Google pays for the default — and the same playbook that has reliably extracted enormous rents from whoever wants the spot.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
