Your iPhone is about to let you pick which AI does the thinking. According to multiple reports published between March and May 2026, Apple plans to let users replace the default AI model behind Siri, Writing Tools, and Image Playground with alternatives from Anthropic (Claude), Google (Gemini), and potentially other providers when iOS 27 arrives later this year. If the plan holds, it would transform Apple Intelligence from a system tethered to a single outside partner into something closer to how you already choose a default browser or email app: your device, your pick.
What the reports say
Bloomberg reported in early May 2026 that iOS 27 will allow users to select third-party AI model providers to power Apple Intelligence features across iOS, iPadOS, and macOS. The scope is broader than many expected: rather than confining outside AI to a chatbot window, Apple would let rival models drive the same built-in tools that currently run on Apple’s own on-device models or route queries to OpenAI’s ChatGPT.
That report built on an earlier Bloomberg piece from March 2026 describing an “Extensions” framework designed to let third-party assistant and chatbot apps integrate directly into Siri, with Claude and Gemini named as expected participants. Reuters also covered the development, though its report was based on Bloomberg’s reporting and unnamed sources rather than independent confirmation.
For readers unfamiliar with the names: Claude is built by Anthropic, a San Francisco-based AI company known for its focus on safety research. Gemini is Google’s flagship AI model family, already available through Google’s own apps on iPhone. Both would join OpenAI’s ChatGPT, which has been integrated into Apple Intelligence since late 2024, as selectable options under the reported system.
What this would actually change for users
Today, Apple Intelligence works as a two-tier system. Lightweight tasks like summarizing a notification or suggesting a text reply run on Apple’s own on-device models. More complex requests, such as asking Siri to draft a long email or generating an image in Image Playground, can be routed to Apple’s Private Cloud Compute servers or, with user permission, to OpenAI’s ChatGPT.
Under the reported iOS 27 plan, that second tier would no longer be locked to a single provider. A user who prefers Claude’s writing style could set it as the default engine for Writing Tools across every text field on their iPhone. Someone who finds Gemini stronger at visual tasks could route Image Playground prompts through Google’s model instead. The change would apply across iPhone, iPad, and Mac, spanning the full Apple Intelligence feature set rather than just Siri.
Separately from the iOS 27 reports, Apple’s existing developer documentation shows that third-party apps can already integrate structured actions with Siri through its App Intents framework. That framework is not the same as the reported “Extensions” system described by Bloomberg, but it demonstrates that Apple has built infrastructure for third-party participation in system-level features. The current ChatGPT integration further shows that Apple Intelligence can hand off queries to an outside model while maintaining system-level privacy controls, including IP address masking and a policy, per Apple’s published disclosures, under which OpenAI does not store requests routed through the system.
The big unanswered questions
Apple has not confirmed any of this publicly. Every substantive claim traces to anonymous sources, which means the timeline, the final list of launch partners, and the technical specifics of the Extensions framework could all shift before a formal announcement. That announcement would most likely come at WWDC, Apple’s annual developer conference typically held in June.
Privacy is the most consequential open question. Apple’s current arrangement with OpenAI includes specific commitments: IP addresses are obscured, and request data is not retained. Whether Anthropic, Google, or other providers would accept identical terms has not been reported. The stakes are not abstract. A user choosing Claude for Writing Tools would be routing potentially sensitive text, from personal emails to medical notes, through servers Apple does not own or operate. How Apple enforces a standardized privacy contract across multiple providers will likely determine whether users trust the system enough to switch away from the default.
Device compatibility is another gap. Current Apple Intelligence features require recent hardware (iPhone 15 Pro or later, M-series iPads and Macs), but no reporting on the iOS 27 expansion has addressed whether the same requirements will apply to third-party model selection or whether additional constraints might come into play. Readers planning to use this feature should watch for hardware and OS details when Apple makes a formal announcement.
Other questions also remain open: whether third-party models will work only via cloud or could eventually run on-device, how model-swapping might affect response quality and speed across different features, and what regulatory considerations apply, particularly around data handling under the EU’s AI Act and GDPR. The financial dimension is also unclear: Apple’s existing deal with OpenAI reportedly involves revenue-sharing arrangements, and whether similar terms would apply to new partners has not been disclosed.
Why this matters beyond the tech
Apple has historically maintained tight control over the core software experiences on its devices. It took regulatory pressure and years of public debate before the company allowed users to change their default browser on iPhone. Voluntarily opening the AI layer to competitors would represent a philosophical shift, not just a feature update.
The two-report arc from Bloomberg, with the second piece expanding scope and adding detail months after the first, suggests a plan that has solidified inside Apple rather than an early-stage concept that might be shelved. Bloomberg’s track record on pre-announcement Apple reporting lends further credibility. Still, until Apple speaks publicly, the specifics remain provisional.
How Apple balances openness with privacy will define the outcome
What is not provisional is the direction. Apple appears to be betting that giving users a choice of AI providers will strengthen its platform rather than weaken it, turning the iPhone into a neutral stage where AI companies compete on quality rather than distribution deals. For the hundreds of millions of people who use these devices daily, the practical question will come down to whether Apple can open that stage without compromising the privacy guarantees that justified trusting Apple Intelligence in the first place.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
