Apple Private Cloud Compute Security Guide

This document describes how Apple Intelligence protects your data when it processes your requests on Apple's servers. Your data is used only to answer your specific request and is not kept afterward. The system is built so that neither outside attackers nor Apple's own staff have a technical path to single out or retrieve your personal data.

This document establishes the security and privacy architecture of Apple Private Cloud Compute (PCC), Apple Intelligence's server-side processing system. It mandates that user data be used exclusively for fulfilling the user's specific request and must not be retained once processing is complete. The architecture is structurally designed without privileged interfaces that could expose user data or allow even Apple's own site reliability staff to bypass privacy guarantees. Non-targetability is enforced as a system-level property: an attacker cannot access a specific user's personal data without attempting a broad compromise of the entire PCC system. External verifiability is built in as a requirement, mandating that security researchers must be able to confirm, with a high degree of confidence, that PCC's privacy and security guarantees match Apple's public commitments.

When you use Apple Intelligence features that rely on server-side processing, your personal data is used only to fulfill your specific request and is not retained after that request is complete. The system is architecturally designed so that no one — including Apple staff — has a privileged interface to access your data, and an attacker cannot target your data individually without attempting to compromise the entire system. A cryptographic method is used to retrieve relevant server-side records without disclosing which records you accessed, preserving your privacy even beyond the core processing boundary. No specific user opt-out or preference action is established by these provisions.