Apple opens private generative AI APIs to iOS and macOS developers

AI · 5 min read

At this week’s developer drop, Apple invited a limited group of partners to test new generative APIs that run primarily on-device using Apple silicon accelerators. The SDK exposes image and text synthesis endpoints, multimodal prompts that mix camera captures with text, and an inference controller that lets apps choose between local or privately federated cloud runs based on user settings.

Apple emphasizes privacy controls: user prompts, model traces, and generated artifacts can be restricted from leaving the device unless the user explicitly opts into cloud processing. There are built-in user-facing disclosure templates and a new entitlement to declare when models might use personal data for personalization.

For designers and product teams, the biggest change may be in the affordances Apple recommends: system-level guardrails to prevent hallucinations in critical UI flows and new accessibility checks for generated visuals. Developers in the beta report smooth integration with Swift and Objective-C, plus sample UI components that match native controls and transitions, shortening the path to prototype-quality AI features.