Apple reportedly designs Siri to delegate tasks to third-party AI models including Claude and ChatGPT

The uncovered frameworks show Apple has built two distinct pathways for third-party AI integration. Model Delegation lets external models act as Siri extensions, handling specific tasks like CSV generation that Siri’s native capabilities do not support. The model then returns results through Siri’s interface, maintaining the illusion of a single assistant while offloading complex requests. This approach keeps Apple’s control over system access but expands Siri’s functional range without requiring Apple to develop every feature in-house.

The more radical Inference Providing protocol goes further by allowing a complete replacement of Apple’s server-side Siri model. In this mode, an external model receives Apple’s planner prompt and tool definitions, enabling it to request system actions like email searches or message sends. The external model processes the data and formulates responses, which Siri then presents using its native voice and UI. This suggests Apple has decoupled its conversational interface from its inference backend, creating a pluggable architecture that could accommodate future AI models without redesigning the user-facing components.

The timing and technical constraints suggest this is a preemptive response to regulatory pressure rather than a user-driven feature. The European Union’s Digital Markets Act likely influenced the design, as it requires Apple to provide third parties with effective access to iOS features, including Siri. However, the current implementation remains internal and undocumented, with no public APIs or entitlements available to developers. This limits immediate utility but signals Apple’s engineering direction, one where Siri becomes a platform for AI interoperability rather than a walled garden.