TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
An Eighth Rank report analyzing macOS 26 says two Apple system processes build a personal knowledge graph from data available on the Mac, including contacts, photos, calendar entries and communication metadata. The report describes relationship inferences, including family-related signals, but Apple’s public documentation is limited and the analysis does not establish every detail of the system’s operation.
Security researchers at Eighth Rank have published an analysis of an on-device personal knowledge graph in macOS 26, describing system processes that connect information about a Mac’s owner, their contacts and their activities. The report says the graph can include inferred relationships among people, but Apple’s public documentation offers only a brief description of the system, leaving details about its inputs and inferences difficult to verify independently.
The report identifies two processes, knowledgeconstructiond and intelligenceplatformd, working together to build the graph. Apple’s macOS man page describes knowledgeconstructiond as analyzing on-device content to build “a general purpose knowledge graph” representing the device and its user. Eighth Rank says it examined macOS 26.6.2 and found the processes’ data in SQLite databases under the user’s IntelligencePlatform directory.
According to Eighth Rank, the directory contains 51 databases, with graph.db storing records that connect entities through defined relationships. The researchers describe a vocabulary of 181 entity classes and 278 predicates. Records can carry confidence, source and timestamp information. The report says sources include Contacts, Photos, Calendar, Family Sharing, Health’s Medical ID emergency contact, and suggestions drawn from Mail and Messages. It also describes separate tables for communication metadata, location and device signals.
The researchers say the communication log records who communicated, when, through which app and in which direction, rather than message content. Other recorded signals include which people appear together in photos and device activity such as screen locking, Wi-Fi, Bluetooth and charging. Eighth Rank also points to a data stream whose fields list people by name, phone number or email alongside time ranges; the researchers say its format appears to track who was physically nearby, though the source material does not establish exactly how Apple uses those records.
A Map of Personal Relationships
A graph that links contacts, events and behavioral records can make scattered information easier for Apple software to use as a connected account of a person’s life. Eighth Rank says the store is queryable and may be used by language models, but that use is presented as an allegation in the report, not as a fully documented Apple description. The distinction matters: a system that organizes data locally is different from a confirmed account of what applications or models retrieve from it.
The report’s findings also matter for privacy and device forensics. A consolidated store may expose patterns that are not obvious from any single message or calendar entry, including associations among people and locations. Eighth Rank says the databases are stored in plaintext within a directory protected by Apple’s Full Disk Access controls. That characterization points to access restrictions as part of the security picture; it does not, on its own, show that the data is exposed to ordinary apps or accessible to an attacker.
Products Worth Considering
Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Midnight
Amazon Kindle Scribe (64GB) with Premium Pen, Metallic Jade
Sony WH-1000XM4 Wireless Noise Cancelling Headphones, Black
As an affiliate, we earn on qualifying purchases.
What the macOS Documentation Says
Apple’s man page, cited by Eighth Rank for macOS 26.6.2, says knowledgeconstructiond analyzes content on the device and works with intelligenceplatformd on heavy processing tasks to generate a knowledge graph. The report argues that this official description says little about the precise data sources, versions in which the components operate, or how particular relationships are inferred.
Eighth Rank notes that Apple researchers published a paper in 2023 on techniques for constructing knowledge graphs, but says that paper does not explain how the system is integrated into macOS or document the kinship inference discussed in the new analysis. The report also points to earlier independent work: researcher Nic Lydon wrote about finding a personal knowledge graph on a Mac, and forensic researcher Alexis Brignoni described examining a related store on iOS in September 2026. Eighth Rank says its own analysis was carried out on macOS 26 and introduces a local dashboard, iKnow, for viewing the graph.
“A daemon that analyzes content on the device to build a general purpose knowledge graph.”
— Apple’s macOS man page for knowledgeconstructiond, as quoted by Eighth Rank
Limits of the Published Evidence
The report is an independent analysis of one macOS release, version 26.6.2, and the supplied material does not include an Apple response. It is not clear from the available documentation which findings apply to other macOS versions, whether every listed data source is used in the same way for every user, or what precise rules determine an inferred family or other relationship.
Eighth Rank says the graph is held in a queryable store that language models retrieve from “allegedly”; the source material does not independently establish which models access it, under what conditions, or whether its contents leave the device. The report’s interpretation of the physical-social stream is also qualified: its record format appears to describe nearby people, but the researchers do not establish the system’s exact purpose or accuracy. Confidence values and source records may help explain individual graph entries, but the article material provides no broad measurement of how often those inferences are correct.
Further Disclosure and Verification
Eighth Rank has released the iKnow dashboard as a local tool for users to inspect their personal knowledge graph, according to its report. Readers should treat the tool and the researchers’ interpretation as independent reporting rather than an Apple explanation of the feature. The next useful developments would be Apple documentation describing supported operating-system versions, data sources, retention and access, as well as a response addressing the claimed relationship inferences.
Further examination across macOS releases and user configurations could show which records are consistently present and how their sources are represented. Until those details are published, the confirmed picture remains narrower: Apple documents a process that builds an on-device knowledge graph, while Eighth Rank’s analysis provides a more detailed account of databases and data categories on one tested system.
Key Questions
What did Eighth Rank find on Macs?
The researchers say macOS 26.6.2 contains processes and SQLite databases used to build an on-device knowledge graph connecting information about the device and its user.
Does the graph include message contents?
Eighth Rank says the communication log records metadata such as participants, time, app and direction, not message content. The report also says suggestions from Mail and Messages can contribute people or events to the graph.
Does Apple confirm that the graph classifies family members?
Apple’s cited man page confirms that a process builds a general-purpose knowledge graph, but it does not describe family classification. The claim about kinship-related inferences comes from Eighth Rank’s analysis.
Can language models access the graph?
Eighth Rank says the graph is a queryable store that language models retrieve from, but the report characterizes that use as alleged. The supplied information does not establish which models access it or under what circumstances.
How can Mac users inspect the graph?
Eighth Rank says it released iKnow, a local dashboard intended to let users read the personal knowledge graph on their Mac. Its report does not establish that the tool is an Apple-supported feature.
Source: 8rank.ai
