TLDR
Hide sensitive content before you share a document. The new Redaction module finds personal and sensitive information in your files and lets you black it out. Before you download, it points out anything that still needs a second look.
See whether your Microsoft labels actually protect your data. Data & More for Purview compares the sensitivity labels in Microsoft 365 with what Data & More has found in your files and emails. It shows where labels are missing or wrong, and can prepare a fix for you to approve.
Get deleted files back. Guided Restore helps you recover deleted Microsoft 365 documents.
More places are covered. Box and Microsoft Teams chats can now be scanned.
Clearer, more reliable results. Classification can now explain why an item was flagged, and scanning runs more reliably across email, file shares, SharePoint and Slack.
You may need access first. Redaction is not switched on for anyone by default. Ask your administrator to give you the Redaction Management permission.
Release date: October 2026 | Previous version: v26.6.11 (13 October 2026)
Scope: Changes since v26.6.0, including what shipped in the v26.6.1 – v26.6.11 patch releases. The headlines are Redaction and Data & More for Purview 2.0. This note also covers new sources, DLP for Teams, LLM-based classification and a broad set of ingestion reliability fixes.
Action required after upgrade. Redaction and Data & More for Purview each come with a new permission. Admin and superadmin users receive DAM for Purview automatically through a new Purview role. Redaction Management is not granted to anyone by default, including Adminlite, so assign it to the people who should redact documents. Data & More for Purview reads sensitivity labels through your Microsoft 365 app registration: if the registration is missing or lacks the label permission, the dashboard says so and shows what to grant.
Redaction
A new Redaction module for removing sensitive content from documents before they are shared or disclosed
Send files straight from a report to Redaction, one at a time or several at once, or upload documents directly
Work is organised in projects with folders and a file tree; files sent from reports and uploads wait in an Inbox until they are moved into a project
Supports PDF and images (PNG, JPEG, GIF, TIFF, BMP, WebP, HEIC); Office and OpenDocument files, email (EML), HTML and text are converted to PDF for review and download
Exchange mail can be sent to Redaction from a report; Outlook .msg files cannot be uploaded directly
Sensitive content is detected automatically and listed as findings that can be reviewed one by one, redacted in bulk or kept
Click individual words to redact them, draw an area by hand, or search the document and redact every match in one step
Areas that could not be confirmed as redacted are outlined in red and flagged "Check redaction" before the file is downloaded
Only one person edits a document at a time; others see a read-only copy with the name of the current editor
Finished files can be reopened, archived and restored within the retention window, or deleted permanently
DLP notification emails and justifications mask values that are labelled for redaction
Data & More for Purview 2.0
A new Purview dashboard in the main menu shows how well Microsoft Purview sensitivity labels protect the data Data & More has classified
Scope shows how many documents are monitored, assessed, classified by Data & More and labelled in Purview
Set Business Logic maps each Data & More classification to the MS sensitivity label it should carry, with optional alternative labels and default labels for files and emails with no sensitive content
Label ranking can be adjusted by nesting labels, and labels that actually enforce protection (encryption, rights management) are marked as protection policies
Measure shows a Purview protection score next to the optimal score a labelling policy could reach, a data protection analysis, and an exfiltration index of sensitive mail sent to external recipients
Detailed Analysis breaks every assessed item down into correctly, over- or under-labelled, not labelled, not mapped, or not labellable by Purview, by label and by classification
Repair generates a Data & More labelling policy that applies the right label to the items carrying the wrong one, modelled on one of your existing policies; it is created for review and does not run until approved
Change Over Time follows coverage and protection through a daily snapshot per workspace
The shared-data chart can be filtered by source: email, OneDrive, SharePoint, Teams, Google Drive or Slack
Reports can show MS sensitivity labels as a column (paired with a policy setting), and "MIP" and "Purview" labels are now consistently called MS sensitivity labels
New features
Sources
Box is available as a source, with its own settings page, accounts and a tagging switch on the policy; enforcement and validation run against Box
Teams chats are ingested, with scan options on the source; a chat appears under every participant on the Data Owners page
MailStore sources can scan only selected mailboxes, and the archive layout is configurable
Rights-protected (MIP-encrypted) Microsoft 365 files can optionally be decrypted for classification
The external Microsoft 365 connect page is redesigned and shows the company and the person who sent the invite
Guided Restore
A new Guided Restore page restores deleted Microsoft 365 documents, behind the Restore Deleted Data permission
Classification
Document classes can use an LLM for classification and can add a written justification to each classified item
Stored justifications can be translated
Document classes can be edited in the client
Norwegian dependency patterns and an updated Danish ethnic-origin pattern
Changes
DLP
Teams messages are classified with the preceding chat messages as context
Teams chat scanning can be limited to tenant groups
Teams deletion consent uses Microsoft's v2 admin-consent endpoint
Content can be classified on OCR labels, faces, signatures and LLM gate classes
Subscriptions left behind by a previous app registration are cleaned up
The DLP policy notification email has a new layout, with dark-mode support and a single restore/review button
Data Owners
Copilot, Exchange (on-premise) and Slack channel sources appear on the Data Owners page
Accounts and SharePoint sites can be copied to the clipboard when grouping by accounts
Users can be assigned as custom owners in the permissions dialog
Assigned custodians show correctly in the report sidebar, and assigning a custodian to a folder no longer propagates up the tree
Reports and dashboards
Mail-based documents show From and To in reports, and email previews no longer show raw mail headers
Policy subscriber reports include every participant of a Teams or Slack chat
Policy notifications can include a help video
Notification and monthly report schedules use your local browser time
Report table: consistent alignment and sorting, hidden column labels no longer stretch the header, and the preview maximise button no longer covers the text
Charts were updated: clickable treemap labels, smoother resizing and modal heights, reworked pie-chart label clicks
Monthly reports that failed on large workspaces now complete
SharePoint sites-at-risk calculation fixed
Sources and ingestion
Ingestion status is labelled Processing or Idle
Starting or editing a source returns immediately instead of waiting up to five minutes
Deleting a source removes its data from the index
Exchange: shared-mailbox ingestion is off by default, account sync no longer wipes sources on start, and an unreadable server response is reported clearly instead of retrying indefinitely
Fileshare: one failing source no longer stops ingestion for all fileshare sources, file names refresh on re-ingest, and a missing share fails once instead of once per file
Graph: a failing message is no longer redelivered for ever, and unresolvable SharePoint sites no longer fail every delta cycle
Slack: failed tasks, delta runs or interruptions no longer stall a source or shrink documents
Items missed earlier are un-purged before items are marked for purging
File processing is shared across Graph, Google, Exchange, IMAP and fileshare sources, so text extraction behaves the same everywhere
Maximum text length is read from global settings
OCR
New OCR engine with layout detection, and a faster route for passports and ID cards (MRZ)
Known persons
Known-person enrichment runs as a per-company job; the CPR and spaCy extraction switches were swapped and now do what they say
A spaCy patterns switch was added to data subject extraction settings
Monitoring
Slow-ingestion alerts now fire for fully ingested sources whose delta runs have stopped, not only during the initial backfill
Alerts are raised on fatal log signatures, and when Elasticsearch is unreachable
Security
Apache Tika upgraded to 3.3.2 across all extractors, closing temporary-file leaks
Spring Boot upgraded to 4.0.8 in the website source
MCP key authentication is accepted on more routes