Apple routes Siri to partner models

Apple is moving Siri closer to a broker model for artificial intelligence, with reports indicating the company plans to send more user requests to outside AI providers rather than relying only on its own systems. The shift would extend a setup Apple already uses with OpenAI’s ChatGPT inside Apple Intelligence. On Apple’s own product pages, the company says ChatGPT is integrated into Siri, Writing Tools, visual intelligence, Image Playground and Shortcuts. The result is a larger role for the iPhone as a distribution point for third-party AI services rather than just Apple-built models. ### How far has Apple already gone with outside AI inside Siri? Apple has already put OpenAI inside core Apple Intelligence features. On its Apple Intelligence page, the company says ChatGPT can work with Siri and other tools “when it might be helpful,” and Apple support documents say Siri can use ChatGPT for “more in-depth answers” on some requests. Apple’s support and privacy materials also show the arrangement is optional and governed through settings. (apple.com) Apple says users can enable or disable the ChatGPT extension, and its legal page says requests are analyzed to determine whether a user is explicitly asking ChatGPT to process them. That means the current system is not a blanket handoff of Siri traffic, but a defined extension inside Apple’s own interface. ### What changes if Siri starts routing to multiple partner models? A broader partner-model system would make Siri less of a single assistant and more of a switchboard. If Apple expands beyond the current ChatGPT arrangement, the company would be deciding which outside model handles which task while keeping the request inside Apple’s software layer, according to the reported direction described in the briefing. That would leave Apple controlling the user entry point even when another company supplies the answer engine. (support.apple.com) Apple signaled in 2025 that it was willing to open parts of its AI stack beyond first-party use. In a June 2025 newsroom post, the company said it was giving developers direct access to the on-device foundation model that powers Apple Intelligence, while also keeping ChatGPT access integrated in Siri and Writing Tools. ### Why does that matter to OpenAI and other AI companies? (apple.com) OpenAI and its backers are not limiting themselves to apps. WIRED reported on June 3 that Opal, backed by OpenAI and Samsung, is working on an AI-powered audio gadget after pivoting from webcams into broader consumer electronics. The report said Opal received a $40 million Series B investment from OpenAI. (apple.com) A separate June 2 report said OpenAI Chief Executive Sam Altman is backing Alfred, a startup founded by former Tesla designer Ankit Ukil and former Meta Reality Labs engineer Dömötör Gulyas. Benzinga, citing Business Insider, said Alfred is developing “physical AI” and is seeking funding at a $40 million valuation. (wired.com) ### Why is the iPhone still the key piece of this fight? Apple controls the device millions of users already carry, and that gives it leverage even when it uses outside models. Apple’s own materials emphasize that ChatGPT can be reached inside Siri and Writing Tools “without needing to jump between tools,” which keeps the interaction inside Apple’s interface and settings structure. That control matters as AI companies test hardware of their own. (benzinga.com) If OpenAI-backed or Altman-backed devices gain traction, they could offer another consumer entry point for AI services. If they do not, Apple still holds the main front-end on the iPhone, where partner access can be granted, limited or expanded through Siri and Apple Intelligence. That conclusion is an inference from Apple’s current integration design and the parallel hardware activity reported around OpenAI-linked ventures. (apple.com) ### What should readers watch next? WWDC has been Apple’s main venue for Apple Intelligence updates, and Apple’s June 2025 newsroom release used that cycle to outline new capabilities and developer access. Any expansion from a single ChatGPT extension to a broader partner-model framework would most likely surface through the same software and platform update process. (apple.com) OpenAI-linked hardware is also moving on a separate timeline. WIRED reported that OpenAI’s own first hardware product is rumored to resemble a smart speaker and could launch in early 2027, while Opal is already developing its audio device. Those next product disclosures will show whether AI companies keep relying on the iPhone for distribution or press harder to build their own hardware channels. (wired.com) (apple.com)