Every Mac comes with an (undocumented) family member classifier
Summary
The article analyzes Apple's Personal Knowledge Graph (PKG) on macOS, describing two on-device daemons that build a graph of a user and their surrounding network using data from Contacts, Calendar, Photos, Messages, and more. It details the data model (graph.db with 181 entity classes and 278 predicates), how edges carry confidence scores, and how inference relies on embeddings, XGBoost classifiers, and on-device ML. It also introduces iKnow, a local dashboard to read the graph, discusses privacy implications and potential attack surfaces, and notes macOS 27 changes; the piece emphasizes transparency and utility versus privacy risk.