Glossary
Definitions of the terms you’ll meet across Enprivacy, in alphabetical order. Each entry links to the page where the concept is covered in context.
Analysis
Section titled “Analysis”The record of one detection pass over a piece of content. An analysis runs the detectors chosen by an analysis plan and produces markers for every sensitive span found. An item’s analyses are listed on its Analyses tab. See The Ingestion Flow.
Analysis Plan
Section titled “Analysis Plan”A named recipe for what to detect: which categories are looked for, which detectors run, and how confidence is scored. Analysis plans are defined estate-wide and are read-only from inside a workspace. Each workspace pins default plans for chats and deep scans — for example the estate Lite plan is the default chat analysis plan. Compare Transformation Plan, which decides what happens to what was detected.
Analyzer
Section titled “Analyzer”One component in the estate’s detection catalogue — a detector that finds sensitive spans (regex, rules, machine-learning models, LLM agents, gazetteers), a refiner that post-processes detections, or a coreference resolver that clusters mentions into entities. Analysis plans bind analyzers together, pinned to the versions in force when the plan was last saved. See Estate: Detection & Governance.
API Key
Section titled “API Key”A workspace-owned key for calling that workspace programmatically (plus registered vendor pass-through keys). Managed under Manage → Connections; keys are revoked rather than deleted so their history stays auditable. Distinct from credentials, which are outbound and estate-wide. See Working in Workspaces.
Artefact
Section titled “Artefact”An output produced from an item by the ingestion pipeline — most importantly the redacted or transformed rendition of a document. Artefacts appear on the item’s Artefacts tab and are stored in the workspace’s redacted repository. See The Ingestion Flow.
Audit Event
Section titled “Audit Event”An immutable record of one action on the platform — a sign-in, a permission change, a chat starting, a message posted. The full trail is browsable by administrators under Admin → Audit Logs.
Category
Section titled “Category”A definition of one kind of sensitive data — a class of PII, commercially sensitive information, or legal material. Categories are defined estate-wide and drive detection: markers are always marked as some category. Workspaces can re-weight how categories score via category scoring.
Category Scoring
Section titled “Category Scoring”A workspace’s per-axis overrides on top of the estate-wide category defaults — how heavily each category weighs in this workspace’s risk scores. Managed under Manage → Detection → Category scoring, with a per-category Reset to the estate default. See Working in Workspaces.
A conversation with the AI assistant, scoped to one workspace. Every message is scanned and redacted per the workspace’s chat analysis and transformation plans before it reaches the model. Chats are private to their owner. See Chats and the AI Assistant.
Credential
Section titled “Credential”An outbound secret for reaching an external system — an AI provider API key, a database password, an OAuth grant. Credentials live in the estate-wide catalogue and are shared by every workspace; a workspace selects which credential its AI chats use via a provider link. Compare API Key (inbound, workspace-owned).
Deep Scan
Section titled “Deep Scan”A heavier background re-analysis. After each chat turn, a deep scan re-reads the conversation with a stronger detector chain (the workspace’s deep-scan plans) and additively redacts anything the fast pass missed — including in earlier turns. See Chats and the AI Assistant.
Entity
Section titled “Entity”A named real-world identity — a person, hostname, account — resolved across the whole estate, so the same person is recognised wherever they appear. Entities are estate-wide and read-only from inside a workspace (Manage → Detection → Entities).
Estate
Section titled “Estate”The organisation-wide layer beneath all workspaces, and also the System area where it is managed. The estate owns the shared catalogues — content sources, database hosts, credentials, categories, entities, analysis plans, and the AI model catalogue. Workspaces import from these catalogues via links rather than owning copies. See The Estate.
Host Link
Section titled “Host Link”A workspace’s import of a database host from the estate catalogue. The link scopes what the workspace sees — the host’s schemas, tables, columns, sample sets, and snapshots — and can be removed without affecting the host itself. The database twin of a source link. See Working in Workspaces.
One piece of content within a source — a document, an email, a file. An item’s page shows its original content, its analyses, its transformations, its artefacts, its tags, and a provenance graph. See Working in Workspaces.
Library
Section titled “Library”A System area holding shared reference books available across workspaces. A workspace imports the books it needs via book links (Manage → Data → Library books).
Marker
Section titled “Marker”One detected sensitive span: this range of this content looks like this category, with this confidence. Markers are what analyses produce, what transformation plans act on, what the inline chat highlights show, and what risk reporting counts. See The Ingestion Flow.
Model Link
Section titled “Model Link”The record that makes an AI model usable in a workspace — in the product’s words, “the chat-creation clamp — without one, chats against that model can’t be started.” Managed under AI → Models. See Working in Workspaces.
Organisation
Section titled “Organisation”A legal entity operating the platform — a controller, joint controller, processor, or sub-processor — recorded in the Estate with its registration details and Data Protection Officer contact, for ROPA Article 30 reporting. See Estate: Detection & Governance.
Permission
Section titled “Permission”A scoped domain:action string (for example workspace.chat:create) from the platform’s read-only registry. Permissions are only ever granted by bundling them into roles. See Admin Workspace.
Provider Link
Section titled “Provider Link”A workspace’s link to an AI provider, carrying the workspace’s default credential and plans for that provider. Managed under AI → Providers. See Working in Workspaces.
A named bundle of permissions. Assigning a role to a team grants every member the bundled permissions in the workspaces that team reaches. Roles exist once platform-wide — editing one changes what every team using it can do. See Admin Workspace.
Source
Section titled “Source”A content system registered in the estate catalogue — a mailbox, an object-storage bucket, an internal repository. Sources are estate-wide; a workspace pulls one into its own relevance set by importing it as a source link. Sources contain items.
Source Link
Section titled “Source Link”A workspace’s import of a source from the estate catalogue. The link is the workspace’s scoped view of that source — its authorisation grant and the path under which its items, analyses, and artefacts appear. Many workspaces can link the same source, and removing a link never touches the source itself. See Working in Workspaces.
Super Admin
Section titled “Super Admin”An account flag (not a role) granting unrestricted access to every surface, bypassing all permission checks. Only a super admin can grant it, never to themselves, and changes are audited. See Admin Workspace.
System Areas
Section titled “System Areas”The platform-wide areas listed under System in the workspace picker, alongside your workspaces: Admin (identity and authorisation), Estate (the shared data catalogues), and Library (shared reference books). Which tiles you see depends on your permissions.
A label attached to content for organisation and search. Tag distribution for a workspace is visible under Manage → Labels → Tags.
The unit access is granted through: users join teams, teams hold roles, and teams are attached to workspaces. A user with no team has no access. See Admin Workspace.
Transformation Plan
Section titled “Transformation Plan”A named recipe for what happens to detected data on its way out: redaction, substitution, or masking rules applied per category. Transformation plans are owned and edited by the workspace (Manage → Plans); first-run setup creates a Default Transformation Plan and a Chat Transformation Plan. Compare Analysis Plan, which decides what is detected.
Transformed Document
Section titled “Transformed Document”The output of applying a transformation plan to an item — the redacted or substituted rendition, stored in the workspace’s redacted repository and surfaced as an artefact. See The Ingestion Flow.
Workflow Run
Section titled “Workflow Run”One execution of a processing pipeline — ingestion, analysis, first-run setup — recorded as a graph of tasks that can be inspected and individually retried under the workspace’s Workflows page. See Working in Workspaces.
Workspace
Section titled “Workspace”An isolated tenant environment — the core unit of work in Enprivacy. A workspace imports sources and hosts, detects and scores sensitive data, applies its transformation plans, and hosts AI chats. Access is granted to teams, never directly to users. See Working in Workspaces.