How Apple’s Personal Knowledge Graph learns about your life on Mac
AI-summarised brief · reviewed before publication
Apple’s macOS 13 introduced a Personal Knowledge Graph (PKG) that aggregates data from messages, emails, photos and other device content to give Siri contextual awareness. Security researchers at Eighth Rank dissected the PKG, revealing two background daemons—named knowledgeconstructiond—that populate 51 SQLite databases in the Library/IntelligencePlatform folder, including a central graph.db with 181 entity classes and 278 predicates. The graph links subjects (people, apps, topics) to objects (names, dates) via predicates that describe relationships such as “has religion” or “face cluster.” Additional tables log communications, location signals, and a “ContextualUnderstanding.PhysicalSocial” list of nearby contacts, though the latter remained empty in tests. The system infers family ties using 144 behavioral features and photo analysis, assigns confidence scores, and retains deleted data in an “expired_stable_graph” for up to 48 hours, allowing derived conclusions to outlive the original records. All data remains stored locally, insulated from cloud exposure.
💡 Why It Matters
- · By keeping such a detailed, inferential model on‑device, Apple offers richer, personalized AI responses while sidestepping the privacy risks of cloud‑based user profiling.