Apple faces a growing dilemma in the artificial intelligence race : how to develop Siri to keep pace with models like ChatGPT and Cloud, without giving up its control over the user experience?
Hot Topics
About half of Libyans drink contaminated water What is the scientific reason?
Mauritania's blue treasure is in danger Will the map of fish in the Atlantic be redrawn?
The Citrus Secret Everyone's Talking About: Burn More, Crave Less & Feel Great Naturally
Indicators in the code of iOS 27 and Mac OS 27 reveal an architecture that may allow for the integration of external AI models with Siri to a deeper degree than the current level of integration.
A software developer and engineer known as pdfu discovered software mechanisms that allow applications to add extensions to Siri, along with indications that the model running behind the assistant could be replaced with an external model, but these capabilities have not yet become a feature available to users, and Apple has not officially announced that it will allow the choice of Cloud or ChatGPT instead of its own model.
Apple's reliance on an external competitor is not new. In 2024, the company added ChatGPT integration to Siri and writing tools within the Apple Intelligence ecosystem, so that the user can benefit from the OpenAI model when its capabilities are appropriate for the task.
However, what appeared in the iOS 27 code indicates a different level of integration. According to developer PDFU, the system includes mechanisms linked to the Model Delegation API within App Intents, allowing an external model to appear as an extension of Siri. A demo showed Cloud running within the Siri interface on a Mac.
This means that Siri’s role may change from a model-based assistant to an interface capable of directing requests to different models, while continuing to manage functions associated with the Apple system.
How can Cloud work within Siri?
In the demo, the user can direct a request to the cloud, which will then understand the request and generate the response. If the request requires an action to be performed within the system, such as creating a reminder, the task can be returned to Siri to be executed via system tools.
This mechanism is consistent with the function of application intent interfaces, which allow applications to define actions and content that Apple's intelligence and Siri can understand and use to perform tasks.
This allows for the separation of two tasks, as an external model handles the reasoning or generating of the answer, while the execution of commands remains within the Apple system and under its supervision.
Can the Siri model itself be replaced?
The most important indicator relates to another mechanism discovered by the developer within Model Manager Services, as published experiments indicate the possibility of replacing the model running on Siri servers with another model.
In a demonstration by the developer, an external model called GPT-5.6 Terra was used to receive Siri prompts and tool definitions, then to handle system actions and obtain their results. The demonstration included summarizing emails and performing an action via the Messages app.
If this mechanism becomes an official feature, it will differ from the current integration with ChatGPT, as the external model will not just be a service that Siri calls on when needed, but could take over a larger part of the reasoning and task execution process.
Tags:
technology
