Apple Turns to Outside AI Models as Siri Falls Behind
Bloomberg’s Mark Gurman says Apple’s reported plan to let users choose outside AI models is a platform move driven partly by weakness in its own technology. Apple aims to make Siri and Apple Intelligence good enough as defaults while allowing services such as ChatGPT, Gemini and Claude to power some features on the iPhone, he argues. Gurman says that could help users in the short term, but it does not remove Apple’s need to build stronger AI of its own for future hardware.
Apple wants the iPhone to host rival AI models because its own AI is behind
Mark Gurman describes Apple’s reported AI shift less as pure neutrality than as an admission of position. Apple, he says, is trying to understand “where it’s at” in artificial intelligence: Siri and Apple Intelligence are “nowhere up to par” with what users see on Android, competing devices, ChatGPT, Claude, and other services.
The reported response is not simply to replace Apple’s AI layer with one outside provider. Gurman frames it as a version of the iPhone’s app model applied to AI. Apple’s default offerings — Siri and Apple Intelligence — need to become “competent and good enough,” the way Apple’s built-in iPhone apps are sufficient for many users. But Apple also wants users to be able to layer their preferred outside AI models onto features, or use those models to power certain functions.
The name of the game is making the default offerings, right, Siri, Apple Intelligence, competent and good enough.
That preserves the core Apple business logic. The customer is still using an iPhone. Apple still earns hardware revenue. And if users subscribe to services such as ChatGPT or Gemini through Apple’s platform, Apple can also take a cut through the App Store.
The “agnostic” posture, then, is bounded. Apple is not giving up control of the device or the platform. It is trying to make the iPhone the place where rival AI services can be chosen and used, while Apple retains the customer relationship and the economics of distribution.
Siri may be rebuilt with Gemini models without becoming Google Gemini
Caroline Hyde points to the current state of Apple’s outside AI integrations: ChatGPT already has a role across some Apple functionality, while reporting indicates Apple is turning toward Alphabet technology for Siri. Gurman distinguishes between two different uses of outside AI.
When Apple Intelligence launched in 2024, Siri was based on Apple models. ChatGPT was available through what he calls an “extension” — a way for Siri or Apple Intelligence to tap into ChatGPT when needed. The reported Siri rebuild is different: Apple is rebuilding its in-house Siri using Google Gemini models, but that does not mean Siri becomes Gemini as a consumer product.
His analogy is that it is “basically like they hired the engineers from Google to fix Siri,” because Google is doing a better job building models. Gemini technology may help power Apple’s assistant, while Apple still presents Siri as Apple’s own interface and experience.
The reported plan has a second track: Apple is moving beyond ChatGPT as the outside service users can invoke. Gemini and Claude are the examples Gurman names. He leaves open whether Apple will allow other providers, naming Meta and Alexa as possibilities rather than confirmed additions.
The distinction matters. One track is infrastructure: Apple using an outside model to make Siri work better. The other is user choice: Apple letting customers select outside AI services to power certain features. Both reduce Apple’s dependence on its own model quality in the short term, but they do so in different ways.
The near-term test is WWDC and a September rollout
Mark Gurman gives a concrete timeline for when users may see the change. Apple, he says, will introduce the new features and the revamped Siri at WWDC on June 8, with rollout in September.
That timing carries stakes because Gurman describes any later rollout as risking “another disaster” for Apple. He adds, though, that he thinks “this is going to happen this time.”
The sequence he outlines is narrow but consequential: announce in June, ship in September, and avoid turning the Siri rebuild into another delayed promise.
Optionality may solve the customer problem before it solves Apple’s strategic problem
Caroline Hyde presses the longer-term question: is this enough, or is Apple deciding to outsource AI? Gurman separates the customer answer from the strategic answer.
For customers in the short term, the arrangement is hard to fault. If users can access multiple leading AI systems from an iPhone, they benefit as Gemini, Claude, ChatGPT, and other services improve. They get optionality rather than being locked into Apple’s weaker default models.
For Apple, the long-term issue is different. Gurman argues that Apple still needs improved models of its own and needs to be “on the frontier” in AI. His reason is hardware: Apple has future hardware plans, and a hardware company does not want to rely on third parties for the underlying AI that powers those products.
That creates the central tension in the strategy. Opening the iPhone to rival AI models can make Apple’s current products more useful and strengthen Apple’s role as the platform through which users reach AI services. But if Apple’s future hardware plans require underlying AI, outside access is not the same as internal capability.