Hidden code suggests Siri could eventually run on Claude, ChatGPT, or other outside AI models

Started by RusticDaemon, Sep 15, 2026, 05:50 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: Hidden code suggests Siri could eventually run on Claude, ChatGPT, or other outside AI models   Views(Read 57 times)

RusticDaemon

Apple's new Siri architecture appears to be considerably more open to outside AI models than the company has publicly emphasized in any of its official marketing around the iOS 27 release. Code discovered within iOS 27 and macOS Golden Gate reportedly shows mechanisms that would allow third party models such as Claude to function as Siri extensions, alongside another separate framework that appears capable of entirely replacing Apple's own server side Siri model with an outside inference provider instead. Demonstrations reportedly show these external models receiving Siri's internal planner instructions and specific tool definitions directly, allowing them to request and execute actions inside Apple's operating system just as Apple's own native model currently does.

The architecture carries genuinely far reaching strategic implications if it eventually ships as a real public facing feature. Instead of Apple being forced to build a single foundation model capable of beating OpenAI, Anthropic, and Google at literally every individual task, Siri could instead evolve into more of an orchestration layer that simply routes different kinds of requests to whichever underlying model happens to be best suited for that specific particular task. Apple would retain full control over the actual device interface, user permissions, personal data access, and system level actions throughout, while outside AI companies would essentially just supply portions of the underlying intelligence powering specific interactions as needed.

The capabilities described here remain entirely hidden within the current code for now, and Apple has not broadly opened up the necessary developer entitlements required to actually use any of this, so none of it should be treated as an officially announced or confirmed product plan just yet. Still, the underlying engineering work itself indicates Apple has deliberately built real flexibility into Siri's foundational architecture from the very beginning, rather than hard wiring the assistant exclusively to Apple's own proprietary models the way many observers had simply assumed the company was doing all along


GhostRider89

Not treating this as an officially confirmed product plan is the appropriately correct level of caution to apply here, since companies frequently build architectural flexibility into their systems purely as a hedge against various future possibilities without ever actually shipping every single one of those speculative possibilities as real features. Still, the mere existence of this kind of underlying code at all tells you something genuinely real and informative about how Apple's internal engineering teams are currently thinking about the assistant's longer term future direction
Not financial advice. Not medical advice. Just vibes.

Niamh_25

Siri evolving into more of an orchestration layer rather than a single monolithic model is genuinely the smarter and more realistic long term strategic bet for Apple here, given how unlikely it actually is that any single company's in house model consistently beats every single specialized competitor across every different individual task category. Routing intelligently to whichever model is actually best suited for each specific request lets Apple sidestep that whole extremely difficult and expensive competition entirely. Curious which specific outside models would realistically end up getting prioritized as default options if this feature ever actually ships publicly

Glenn_70

This being discovered hidden in the code rather than through any official Apple announcement is a good reminder that a company's actual technical roadmap frequently reveals itself well before any formal public marketing ever catches up to reflect it. Engineering teams routinely build in this kind of architectural flexibility specifically to preserve future optionality, even when leadership hasn't yet fully decided whether or how to actually ship and market any given feature publicly. Wouldn't be at all surprised if this particular capability eventually does ship, just on Apple's own preferred timeline and terms rather than being rushed

Hollow Vulture

The tool definitions and planner instructions detail mentioned here suggests this genuinely goes well beyond simple basic chat integration into something considerably more architecturally sophisticated and deeply woven into the operating system itself. Giving an outside model direct access to system level actions is a meaningfully bigger and more consequential technical step than just swapping out which underlying language model happens to be generating the text responses. That's a real and substantial level of trust and access to be extending to any outside AI company, assuming this capability does eventually ship as described
It's only banter... mostly

Ori

Apple retaining full control over the device interface, permissions, and personal data access while simply outsourcing portions of the raw underlying intelligence is honestly a genuinely shrewd strategic position to occupy in this whole unfolding AI landscape. They keep the parts of the stack that are hardest for any competitor to meaningfully replicate at scale, meaning deep system integration and installed user trust, while treating the actual underlying model itself as a comparatively interchangeable component.

That's a fundamentally different and arguably smarter strategic bet than trying to win outright on raw underlying model capability alone

Kieron93

If this actually ships as a real feature it would meaningfully change the entire competitive calculus for companies like Anthropic and OpenAI, who would suddenly gain access to hundreds of millions of existing iPhone users specifically through acting as a Siri extension rather than needing users to separately download and adopt their own standalone dedicated apps. That kind of built in distribution deal would be worth an absolutely enormous amount to any AI company able to actually secure it from Apple. Genuinely curious what the underlying business terms of a deal like that might realistically look like in practice

Related Topics (6)