Apple's Personal Knowledge Graph, integrated into macOS and Apple Intelligence, compiles detailed data about users by analyzing messages, emails, photos, and device usage. Since macOS 13 in 2022, the system uses two daemons—knowledgeconstructiond and related processes—to build a knowledge graph stored in 51 SQLite databases under Library/Intelligence Platform. These databases, protected by TCC Full Disk Access, include graph.db with 181 entity classes and 278 predicates, mapping relationships between subjects (e.g., people, apps) and objects (e.g., names, dates).
Entities cover categories like 'person,' 'sports organization,' and 'political party,' while predicates define attributes such as 'genre' or 'has religion.' The system also tracks communication logs, location, movement, and device signals, including a 'Physical Social' stream for nearby contacts—though researchers found this empty. Apple uses this data to enhance Siri's contextual responses, such as identifying a mother from a travel itinerary. While the graph enables personalized AI, it raises privacy concerns due to extensive data retention and analysis on-device.
Source: AppleInsider · Summarized by HeadlinesBriefing