Mira AI: eDiscovery agent for Documents — Redacting Sensitive Data

When you fulfill a Get a copy / Access request, you often need to remove sensitive information from the files before sending them to the data subject — such as other people's names, employee details, or third-party data. The eDiscovery agent for Documents is an AI agent that scans the files attached to a request, detects the sensitive data types you've configured, and suggests them for redaction, so you only need to review and confirm.

For structured data in databases and SaaS apps, use the eDiscovery agent for SaaS & DB instead.

What the agent does

Scans unstructured files (such as PDFs, Word, PowerPoint, Excel) attached to access requests.
Runs automatically after files are collected by an integration or uploaded manually, so suggestions are ready when you open the request.
Marks the data types you've selected — plus any custom patterns — with red boxes in the document viewer.
Always excludes the data subject's own information (their name, email, address, and phone from the request details), so you never redact the requester's own data.
Only suggests — you review and confirm every redaction.

Availability

The eDiscovery agent for Documents is available on the Enterprise plan. You enable and configure it from the Mira AI agents page.

Enable the agent

On the Mira AI agents page, open the eDiscovery agent for Documents tile and turn the agent on. A confirmation explains that the agent runs after files are uploaded to a request; to apply it to requests that already exist, use Re-run eDiscovery in the request.

Configure which data types to redact

In the agent's Setup, choose the data types the agent should look for. They're grouped into categories (Contact Info & Location, Demographics, Government & National IDs, Financial Info, Health Info, Credentials & Secrets, Device & Network identifiers, Web & Technical Data, and Custom). Tick a whole category or individual types. MineOS starts you off with a default selection of common types (name, email, phone, address, financial identifiers, national IDs, and more). Hover any type to see what it covers.

Click Save to store your selection. The configuration is saved at the company level (not per brand), and only Admins and Owners can change it.

Add a custom pattern (regex)

To catch values that aren't in the standard list — add a custom regex:

Click Add custom regex (under the Custom category or next to Save).
Give it a name and enter the regex pattern.
Click Add.

The pattern appears under Custom, showing its name with a 3-dot menu to edit or delete it. Names and patterns must be unique — MineOS won't let you add a duplicate name or an identical pattern.

Changing the regex configuration does not automatically re-scan requests that were already collected. To apply new regex settings to such a request, use Refresh all in the request's page.

Regex suggestions and pre-defined data types suggestions are separate processes, so if the agent matches a value for both a marked data type and a marked regex pattern, you will see in the UI 2 red bounding boxes.

Redact a request's files

Open an access request, then open a collected file in the redaction view. The left panel shows the customer's details and the list of files.

Review the agent's suggestions

Once the agent finishes scanning a file, a banner confirms it completed scanning this file for data types, and detected values are marked with red boxes. While the agent is still scanning, the banner says so, you will have a spinner indication near the file in the file list, and the Mark button is disabled until the agent finishes the scan.

Detect data types — choose what's shown

Click Detect to open a panel of all data types. The ticked ones match your current Mira AI configuration. Tick or untick types to change which suggestions appear — this is a display filter (the agent already scanned for everything), so it applies instantly without re-running the agent. Click Update to save the selection; it applies to all files in this request, not just the current file.

Mark data manually

Click Mark to draw your own boxes — the page prompts you to click and drag to hide sensitive details. To exit selection mode click on Stop selecting, while in selection mode the upper banner changes and other buttons are hidden.

Each manual selection is added to the total selection count, and you always have an option to Clear all suggestions, just note that this includes both the manual marked suggestions and automatically suggested by the agent.

Remove a suggestion

To drop a suggested box, click on the suggestion itself and choose Remove selection (or Keep selection to keep it).

Apply and save

Click Apply redaction to turn the marked boxes into actual redactions (they become solid black).
Click Save to save the redacted file. Saved files are treated as complete.

Supported file types and size limits

Files are limited to 10MB, whether they're uploaded manually or retrieved automatically by an integration.

Supported file types:

Office — .docx, .doc, .dotx, .docm, .xlsx, .xls, .xlsm, .pptx, .ppt, .pptm, .ppsx, .pps, .rtf, .odt
CSV — .csv, .tsv
Text — .txt, .text, .log
HTML — .html, .htm
Email — .msg, .eml, .pst, .mbox, .mbx
Image — .jpg, .jpeg, .png, .gif, .bmp, .tiff, .tif, .svg