The December 2025 Redaction Failure
Updated for 2026
The US Department of Justice released the Epstein files in December 2025. News coverage quickly shifted from the content to the redactions. It focused on how easily those redactions could be bypassed.
The method was simple. "Redacted" content in PDF files was blacked out with highlighting. But the words stay in the PDF's text layer. Copy the black box into a text editor and the original words appear. The visual cover was not true deletion. The sensitive data was never removed.
This was not a new flaw. The 2007 Anthony Pellicano case had sensitive data revealed through improper redaction in legal papers. The same failure showed up in court filings and government reports for years. Yet the Epstein files made the failure visible to tens of millions of people in real time.
For more on document de-identification in legal contexts, see our compliance overview.
Visual Cover vs. True Redaction
Why does this keep happening? The answer lies in a key technical gap. There is a difference between visual cover and true deletion.
Visual cover places an element over words. It does not remove those words from the file. These methods all fall into this group. Black highlighting sets the background to black. White words on a white page change the color to match. A black rectangle drawn over text hides the view. PDF annotation cover adds an opaque layer on top. An image overlay places a black image over the words.
In every case, the original words stay in the file. They can be found by copying the region or removing the overlay. A person with tech skills can also scan the raw file.
True redaction removes the words from the file for good. The content is not hidden — it is gone. Nothing is left to find.
The key question for any file you send out: when someone checks this file, will they find the original words? With visual cover, the answer is yes. See our glossary for definitions of redaction terms.
The Word Document Problem
The same failure exists in Microsoft Word. Using black highlighting or opaque boxes to "redact" a Word file leaves the original words in the document's XML.
This matters because Word is the main format for legal letters, contracts, HR files, and internal reviews. Groups using highlighting have been sending records with data that can be found throughout their history.
71% of legal teams use AI tools despite information storage concerns (ACC 2025 survey). As AI tools enter document work, the risk of surfacing past redaction failures grows. An AI that reads your files may find words in "redacted" sections that were never actually deleted.
High-Profile Redaction Failures
The Epstein files were not the first high-profile case of this failure.
The Anthony Pellicano case (2007) involved sensitive data revealed through improperly redacted papers filed in federal court. [VERIFIED-EXTERNAL]
NSA documents released via FOIA requests have repeatedly had readable words under black boxes. Security researchers have documented this in national security releases. [VERIFIED-EXTERNAL]
Corporate litigation filings often have readable content when parties use PDF annotation layers instead of true deletion. [VERIFIED]
This pattern shows a basic gap. Legal teams think of redaction as a visual act. But PDF and Word formats contain structured data no matter what you see on screen.
What True Redaction Needs
For a file to be truly redacted, the words must be removed and replaced. A skilled person must not be able to recover them.
In PDF files, true redaction means four things. First, flatten the PDF to remove all editable layers. Second, replace content with black boxes at the content stream level. Third, remove metadata that may hold the original words. Fourth, remove embedded fonts that could allow recovery.
In Word files, true redaction means three things. First, find every instance of the target content — in tracked changes, comments, metadata, and revision history. Second, replace the content, not cover it visually. Third, preserve the format without leaving marks.
The key word is replacement. The original content must be replaced with something else, not hidden beneath something else.
Headers, Footers, and Hidden Zones
Legal document redaction has more layers than just the main body. Sensitive data often appears in zones that visual tools miss entirely.
Headers and footers often contain matter names, client IDs, and document numbers. Blacking out a contract body while leaving "Privileged — Re: TechCorp" in the header defeats the purpose.
Comments and tracked changes are a common source of unintended disclosure. A reviewer who comments "see John Smith's note" leaves that in the file. It stays even after the clause is covered.
Document properties and metadata contain author names and revision history. These can reveal the document's origin even when the body is blacked out.
Revision history in Word preserves prior versions of edited content. A file that once said "the plaintiff's home address is 123 Main Street" keeps that version. It stays unless you clear it.
Building a Compliant Process
Given these failure modes, a sound redaction process needs four steps.
1. Use native Word integration for Word files. Redaction within the Word object model replaces content directly in the file. This avoids the concealment problem. Converting to PDF first adds risk and may miss comments and revision history.
2. Process all document zones. A compliant process must handle headers, footers, footnotes, endnotes, comments, tracked changes, and document properties — not just the main body.
3. Verify the output. After redaction, try to recover the content. Copy-paste the redacted areas. Check the document's XML. Review tracked changes and revision history. If original content appears anywhere, the redaction is incomplete.
4. Maintain an audit trail. For legal productions, record what was redacted, by what method, and by whom. This matters if a privilege dispute arises. Learn more at our security and conformance page.
Lessons from the Epstein Files
The Epstein files failure was a public lesson. It showed what happens when visual cover is confused with true redaction.
Every legal team and compliance pro who watched this story should ask two questions. First, what is in our past document productions that could be similarly recovered? Second, does our current process actually delete content or merely cover it?
The answers determine real exposure — not just the existence of a redaction policy.
anonym.legal's Office Add-in performs true PII replacement within Word files. It replaces content directly in the document structure, not over it visually. Headers, footers, footnotes, comments, and tracked changes are all processed. The result is a file from which the original data is absent, not hidden. Learn more.