Apple's redesigned Siri appears built to support third-party AI models like Claude and GPT-5.6

Developer pdfu, after digging into iOS 27 and macOS Golden Gate, says Apple has designed its new Siri AI architecture in private frameworks to work deeply with third-party AI models. Demo videos show both Anthropic's Claude and OpenAI's GPT-5.6 being integrated into the Siri framework.

One mechanism, called "Model Delegation," lets Claude appear as a Siri extension, similar to the built-in ChatGPT extension. In a macOS demo, users could invoke an "Ask…" menu from the search or ask bar to choose Claude. After enabling the extension, a user asked Siri to "have Claude" create a reminder in the Reminders app; Claude parsed the natural-language request and then handed it back to Siri to perform the system action. In another example, Claude handled a request to create a CSV file inside the Siri conversation window—something Siri itself could not do—and returned a result.

The second protocol, an "Inference Provider" in the "Model Management Services" framework, goes further: it could allow users to completely replace Apple's on-device Siri model with a model such as GPT-5.6. In that mode, ChatGPT would take over Siri's planner and tool definitions, allowing it to request system operations, receive personal data in response, and generate answers that Siri presents through its own interface and voice. In a demo, a user asked ChatGPT to find and summarize emails and to-dos inside Siri, then send a message via the Messages app, with the responses appearing in the OpenAI web dashboard.

Most media analysis links Apple's approach to the EU's Digital Markets Act, which requires Apple to give third parties access to iOS hardware and software features equivalent to its own services. For now, the "Ask…" implementation is limited to the ChatGPT extension in the macOS 27 Golden Gate candidate, with Claude unavailable; Apple has not yet opened model delegation authorization to third parties. Still, the code suggests Apple has done substantial engineering groundwork for future model interoperability.