TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
A report by Eighth Rank describes a macOS service that builds an on-device knowledge graph from information and activity signals across a Mac. Its analysis says the graph includes relationship categories, but the exact kinship inferences, coverage across Mac models and privacy controls remain unclear.
Security research firm Eighth Rank has published an analysis of a macOS 26 on-device knowledge graph that it says combines personal records and behavioral signals to map information about a Mac’s owner and their contacts. The report describes relationship categories in the graph, including family-related information, but Apple’s public documentation does not explain the system’s full data inputs or how it draws kinship conclusions.
Eighth Rank says it examined the system on macOS 26.6.2. It identifies two processes, knowledgeconstructiond and intelligenceplatformd, as working together to build the graph. The macOS man page for knowledgeconstructiond describes it as analyzing content on the device to create “a general purpose knowledge graph” representing the device and user. That description does not spell out the detailed sources or inferences in the report.
The researchers found a set of 51 SQLite databases in the user’s IntelligencePlatform directory and focused on graph.db. They describe it as a triple store, with entities connected by predicates that represent relationships. The report says its vocabulary includes 181 entity classes and 278 predicates; entries also carry confidence, source and timestamp information. The databases are described as plaintext and stored in a location protected by macOS privacy controls, including Full Disk Access restrictions.
According to Eighth Rank, graph inputs include Contacts, Photos, Calendar, Family Sharing, Health emergency contacts, Siri Suggestions data from Mail and Messages, app-provided information and installed apps. Separate tables reportedly record communication metadata, photo co-appearance, location and movement, and device signals. The communication log is described as recording who communicated, when, through which app and in which direction, rather than message content. The report also points to a record format for a physical-social context signal that lists people with start and end times; what that signal means in practice is not independently established by the material provided.
A Detailed Map of Personal Connections
A graph that links contacts, events, communications and location can make information useful across apps and services, but it also concentrates sensitive details about a person’s relationships and routines. Eighth Rank argues that such a store could create a privacy risk if accessed by someone who should not have it. That is a risk scenario raised by the researchers, not evidence that the graph has been accessed or misused.
The report matters to Mac users because Apple’s brief man-page description does not provide the level of detail Eighth Rank says it found through database analysis. Understanding which sources contribute data, what the system infers and how long records persist can affect users’ expectations about privacy and device-level personalization. The analysis does not establish that every listed source is used for every user, or that the graph is available to a particular AI model in all circumstances.
Products Worth Considering
Samsung Galaxy Tab A11 8.7-Inch Wi-Fi Tablet, 64GB, 4GB RAM, Silver
Logitech MX Master 4 Ergonomic Wireless Mouse with Haptic Feedback, Graphite
Logitech MX Keys Combo for Business Keyboard and Mouse – Graphite
As an affiliate, we earn on qualifying purchases.
What Apple Documents Publicly
Apple’s knowledgeconstructiond man page calls the service a device-side analyzer that works with intelligenceplatformd and performs heavy processing to generate a knowledge graph. Eighth Rank says Apple researchers published work on graph-construction techniques in 2023, but it found little public documentation describing how those techniques are integrated into macOS or specifying the full set of data and inferences.
The report is not the first published account of the store. Eighth Rank notes that Nic Lydon had previously written about finding a personal knowledge graph on a laptop, and that forensic researcher Alexis Brignoni later described examining a similar store on iOS using the iLEAPP tool. Those accounts provide related observations, but they do not by themselves establish that the same data collection and inference behavior applies identically across all devices or operating-system versions.
“A daemon that analyzes content on the device to build a general purpose knowledge graph.”
— macOS knowledgeconstructiond man page, as quoted by Eighth Rank
Limits of the Available Evidence
The analysis is based on one macOS 26.6.2 examination, and the provided source does not establish how the system behaves across every Mac, account configuration or operating-system release. Apple has not, in the cited documentation, confirmed Eighth Rank’s complete inventory of data sources or explained how the graph’s relationship categories are populated.
The report’s headline characterization of a family-member classifier should be treated cautiously: the supplied material describes relationship types and references family-related data, but it does not include a detailed account of the specific kinship rules, their accuracy, or the situations in which the system infers family ties. It is also unclear from the material how long records are retained, which Apple features or models can query the graph, and what user controls are available beyond the stated macOS privacy protections. No misuse or exposure incident is reported.
Further Testing and Apple’s Response
Eighth Rank says it released a local dashboard called iKnow for users to inspect their personal knowledge graph. The source material does not provide results from a broad independent evaluation of the tool or establish how its findings vary between Macs, so users should treat any display as a view of the examined device rather than a universal account of macOS behavior.
More detail would be needed to establish the graph’s full reach: testing across different macOS versions and configurations, documentation of relationship-inference behavior, and an explanation of retention and access controls. The supplied report contains no response from Apple addressing its findings. Until such information is available, the specific extent of the family-related classification and any changes Apple may make remain unknown.
Key Questions
What did Eighth Rank report?
Eighth Rank says macOS 26 builds an on-device knowledge graph from personal records and activity signals. Its researchers examined databases and processes on a Mac running macOS 26.6.2.
Does the report prove every Mac classifies family members?
No. The analysis describes relationship categories and family-related inputs, but the supplied material does not establish the exact kinship rules or confirm identical behavior across all Macs and macOS versions.
Does the graph store message contents?
Eighth Rank describes a communication log that records participants, timing, app and direction, and says it does not contain message content. The report separately identifies data sources such as Mail and Messages, so the full handling of those sources is not resolved by that description alone.
Where did the researchers find the databases?
The report places the databases in the user’s Library/IntelligencePlatform directory and says they include SQLite files, including graph.db. It describes the location as protected by macOS privacy controls, including Full Disk Access restrictions.
What happens next?
Eighth Rank has released the iKnow dashboard for local inspection, while broader testing and clearer documentation would be needed to verify how the graph works across devices and versions. The source material does not include an Apple response.
Source: 8rank.ai
