Apple introduces Private On-Device ML APIs for App Store developers

AI · 5 min read

At WWDC follow-up updates, Apple unveiled Private ML, an API layer that standardizes on-device inference across iPhone, iPad and Apple Silicon Macs. Developers can bundle encrypted model assets which are verified and sandboxed by the OS, reducing the need to route requests through remote servers.

The framework includes a privacy-first prompt flow, letting users review what models an app uses and managing a per-app model permissions panel. Apple is also offering a memory-optimized runtime and hardware-accelerated quantization paths for Core ML-compatible models to improve battery usage.

Apple emphasizes the ecosystem benefits: by enabling more on-device features without cloud telemetry, apps can offer personalized experiences while remaining compliant with Apple’s privacy guidelines. The App Store review notes for apps using Private ML will include automated checks to ensure secure model packaging and transparency.