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 describes macOS 26 services that build an on-device knowledge graph from data and signals available to the operating system. The report says its records include relationship information and confidence scores, but Apple’s public documentation gives little detail about the system’s sources or inferences.
Eighth Rank examined the system on macOS 26.6.2 and identified two background services, knowledgeconstructiond and intelligenceplatformd, associated with building and processing the graph. Apple’s man page describes knowledgeconstructiond as analyzing on-device content to build “a general purpose knowledge graph” representing the device and its user. The report says the two services store data in a set of SQLite databases in the user’s IntelligencePlatform directory.
The report identifies graph.db as the central store. It describes a triple-based structure, with records linking a subject to an object through a predicate, and says the graph uses 181 entity classes and 278 predicates. Rows can carry confidence, source, duplicate-source and timestamp information. Eighth Rank says the system draws on sources including Contacts, Photos, Calendar, Family Sharing, Health emergency contacts, and suggestions from Mail and Messages. It also describes separate tables for communication metadata, location and movement, and device signals.
According to Eighth Rank, the communication records identify who communicated, when, through which app and in which direction, but do not contain message content. The report also describes a record stream named ContextualUnderstanding.PhysicalSocial, whose format includes names, phone numbers and email addresses with start and end times. The authors say this appears to record people physically near the user; that interpretation is theirs, rather than a claim established by Apple documentation. Eighth Rank released a local dashboard, iKnow, to inspect a Mac’s graph.
A Personal Map Built On Mac
A graph that connects people, events and activity can provide useful context for on-device features, but it also concentrates information that may be sensitive even when individual records seem routine. Relationships, communication patterns and location signals, taken together, can reveal a detailed picture of a person’s life. Eighth Rank frames this as a privacy concern, particularly if someone gains access to the store.
The analysis also gives users and researchers a clearer account of what may be present on a Mac beyond familiar apps and files. Its description of relationship confidence and source records matters because inferred connections may be mistaken, and because the provenance of an inference can affect how users interpret it. The report does not establish that the graph is routinely exposed or that it has been misused.
Products Worth Considering
Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Midnight
Bose Lifestyle Ultra Wireless Home Speaker
Apple iPad 11-inch (A16 Chip, 128GB, Wi-Fi, Pink)
As an affiliate, we earn on qualifying purchases.
What Apple Documents Publicly
Apple’s man page for knowledgeconstructiond says the service analyzes device content and works with intelligenceplatformd on processing to generate a graph representing the device and user. Eighth Rank says Apple researchers published work on graph-construction techniques in 2023, but argues that public material does not explain how the mechanism is integrated into macOS, its full source list, or specific inferences such as kinship.
The report builds on earlier independent investigations. Eighth Rank says researcher Nic Lydon had written about finding a personal knowledge graph on a laptop, and that forensic researcher Alexis Brignoni later added support for reading a similar store on iOS to the iLEAPP tool. Those reports indicate interest in related data stores across Apple devices, but the Eighth Rank analysis detailed here is specifically based on macOS 26.6.2.
““A daemon that analyzes content on the device to build a general purpose knowledge graph.””
— Apple, in the knowledgeconstructiond man page quoted by Eighth Rank
How The Inferences Are Made
The report is an independent analysis of a particular macOS version, not a complete Apple specification. Apple has not publicly detailed in the material cited which macOS releases run these services, the exact rules used to infer family or other relationships, or how those inferences are surfaced to Apple Intelligence features. Eighth Rank’s interpretation of the physical-social records is also not confirmed by Apple.
The authors say the databases are stored in plaintext within a directory protected by Full Disk Access controls. The report does not establish that any unauthorized party has accessed the graph, nor does the supplied material quantify how often records are wrong or how long particular information is retained. It also does not show that a large language model necessarily retrieves every graph record; the report characterizes that point as alleged.
Further Review Of The Graph
Eighth Rank has made its iKnow dashboard available for local inspection, giving users and researchers a way to examine the graph on their own Macs. The report does not specify a response or product change from Apple. Further clarity would depend on Apple publishing more about the services, their data sources, retention practices, and controls for inspecting or removing inferred records.
Users who want to investigate can review Eighth Rank’s analysis and tool documentation, while keeping in mind that the findings concern the tested macOS build. Whether Apple will expand its documentation, and whether the same schema and inputs apply to other versions or devices, remains open.
Key Questions
What is the Mac knowledge graph?
According to Eighth Rank, it is an on-device store that links information about the user, people, events and other entities. Apple’s man page describes a general-purpose graph built by analyzing device content.
Does the graph include message contents?
Eighth Rank says its communication log records who communicated, when, through which app and in which direction, but not message content. The report separately says Mail and Messages can contribute contact or event suggestions.
Does Apple confirm that Macs infer family relationships?
The report describes kinship-related inference, but the Apple documentation it cites does not explain that capability. The specific claim is based on Eighth Rank’s analysis, not a detailed public statement from Apple.
Can users inspect the graph?
Eighth Rank released a local dashboard called iKnow for examining the personal knowledge graph. Its report says the databases sit in a directory protected by Full Disk Access controls.
Is there evidence the data has been misused?
The supplied report raises a hypothetical privacy risk if the store were accessed by someone unauthorized. It does not provide evidence that the graph has been breached or misused.
Source: 8rank.ai
