Spreadsheets are where DSAR redactions go to fail. A workbook is not what it looks like on screen: it is a container of sheets, some hidden, holding cells, some hidden, plus caches, comments, and properties that never render at all. Redact the visible grid and you may have redacted a fraction of the file. This page covers where personal data hides in a workbook, how detection has to work to find it, and what to check before a spreadsheet ships in a disclosure bundle.
Where Data Hides in a Workbook
Six layers, in rough order of how often they bite. Hidden rows and columns: filtered-out or manually hidden ranges keep their values; a payroll sheet with columns G through K hidden still contains columns G through K. Hidden and very hidden sheets: a workbook lists every sheet internally whether or not the tab shows, and “very hidden” sheets do not even appear in the unhide menu, only in the file itself. Pivot caches: a pivot table remembers a copy of its source data inside the file, so deleting the source sheet does not delete the data. Comments and notes: threaded comments carry names and often the exact content someone thought was safely informal. Document properties: author, last-modified-by, and custom metadata fields carry names by default. And formulas versus values: a cell can display one thing and compute from another sheet you thought was gone.
How SafeRedact Reads a Workbook
Detection operates on the file’s data, not its display state. SafeRedact’s spreadsheet extraction walks every sheet in the workbook, including hidden and very hidden ones, and reads every populated cell, including cells in hidden rows and columns, then runs detection over the cell values and row contexts. Names, emails, phone numbers, addresses, and identifiers in ranges no reviewer would ever scroll to get the same treatment as the visible grid. That is the property to demand of any tool in this workflow: if extraction mirrors what the screen shows, the screen’s blind spots become the bundle’s disclosures.
The Pre-Send Checks Specific to Spreadsheets
Whatever produced the redacted output, verify it the way a motivated recipient would read it. Unhide everything: all sheets, all rows, all columns, and read what appears. Check for very hidden sheets, which need the file inspected rather than the unhide menu. Refresh or delete pivot tables and see what their caches still know. Open the comments pane. Inspect document properties. And run the general clean-view test: export or copy the full content as text and search it for the names and identifiers that were supposed to be gone. Five minutes per workbook, against the cost of a disclosure incident.
CSV: the Honest Cousin
One reason DSAR pipelines like CSV: it has no hidden layers. A CSV is exactly its text, every value visible to any editor, which makes it easy to verify and hard to booby-trap. The trade is that one CSV carries one sheet and no formatting, so a workbook becomes several files. Where the disclosure needs the data rather than the workbook experience, flattening to CSV before or after redaction removes an entire category of failure, and SafeRedact processes CSV natively alongside XLSX.
Why This Matters More Under the 2026 Rules
A spreadsheet that leaks a hidden column of third-party data is a disclosure incident, and from 19 June 2026 the affected people also have the section 164A complaints procedure to raise it through, on the record. For US state requests the same layers matter in the other direction: Minnesota’s confirm-but-not-disclose rule means the requester’s own account numbers must come out of the workbook before it ships. Either way, the hidden layers are where the failure lives.
How SafeRedact Fits
Files never leave your browser: documents are processed client-side, only extracted text is sent for AI detection, and no documents are stored on servers. Detection text is never used for training and is deleted within 30 days. DSAR mode lets you name the data subject whose information should be preserved while other individuals’ personal information is flagged for redaction across every file in the case. Every detection is reviewed by a human before anything is exported, and the output includes a processing summary and audit trail your privacy officer can stand behind if the ICO asks how the response was produced.
A Data Processing Agreement is available, and enterprise cases are priced per case, with the price set with you and shown before you download. See the enterprise overview or the enterprise FAQ for how cases run end to end.
Redact the Workbook, Not the View
Every sheet and every cell read, hidden or not, with human review of every detection and an audit trail of what was removed.
Start Your EvaluationRelated Guides
Microsoft, Microsoft 365, Outlook, SharePoint, Exchange Online, OneDrive, Teams, and Purview are trademarks of Microsoft Corporation. SafeRedact is not affiliated with or endorsed by Microsoft. This page is informational, not legal advice.