The Highlight-to-PDF Trap
Every week, attorneys make the same mistake. They highlight text in black, convert to PDF, and call it done. That content is not hidden. Anyone can select it, copy it, and read it in seconds. It happens routinely. The error has real consequences.
Courts have noticed. Federal judges are not amused.
What Highlighting Actually Does
Highlighting places a colored layer over words. It does not remove anything beneath. The original data stays inside the file. It is searchable, copyable, and accessible to anyone.
Real document scrubbing is different. It removes the underlying data from the file entirely. The visual bar and the data removal are two distinct steps. Most attorneys only do the first one.
Courts Catching the Error
The "Technical Weakness" Warning
A magistrate judge recently demanded that counsel explain. The filing had a "technical weakness" in the redaction process. The attorney had used Word's highlight tool. Opposing counsel selected the covered passage and copied it out.
The Metadata Leak
An attorney produced files where names and dates appeared blacked out. Those same details were readable in the document properties. Metadata had not been stripped. Nothing had been removed from the file.
The Mass Recall
A paralegal found that covered passages could be copied from under black highlighting. The firm had already sent thousands of files in that format. Every document had to be recalled and re-produced. The cost and embarrassment were significant.
Federal Court Guidance
The federal judiciary is direct:
"Merely drawing black lines through text does not securely redact. The underlying text may still be searched, copied, or accessed."
The official guidance has four requirements. Use tools that remove data, not tools that cover it. Flatten PDF files after processing. Verify the output before filing. Strip metadata from every document.
What Proper Scrubbing Requires
These are the technical requirements.
Data deletion. The words must be removed from the file. A colored box over the words is not enough.
Metadata stripping. Document properties can expose the original content. Author name, timestamps, and revision history must all be cleaned.
Layer flattening. This prevents someone from lifting the black box off the passage.
Audit trail. Records what was removed and when. Essential for the privilege log. Courts may ask for this record.
The process has five steps. Review all content that must be removed. Use a tool that deletes the data. Verify the passage is gone from the saved file. Document what was removed and why. Then export the clean copy. Do not overwrite the original.
The Discovery Dilemma
E-discovery adds a complication. You must produce documents. But courts can later order original, un-covered production. Permanent removal creates a real bind.
A judge may order the original version produced. A client may need access to the source document. An auditor may request complete records. An appeal may require the original evidence.
Reversible encryption solves this. Instead of deleting a passage, you replace it with a token. The key stays with your firm. The file goes out clean. If a court orders disclosure, you decrypt with the key and produce the original.
How it works: Reversible Encryption for Legal Discovery.
Related reading: Word Add-in Redaction Costs at Law Firms.
Next Steps for Law Firms
The "technical weakness" excuse no longer works. Courts are more aware of these failures than before. Sanctions are more common now. The standard for attorneys is clear.
Three immediate actions:
- Stop highlighting for any filing or production.
- Add a copy-paste check. Paste the output into a plain text editor. Confirm the passage is absent.
- Strip metadata before every production run.
anonym.legal's Office Add-in handles all three inside Microsoft Word. It removes data, flattens the output, and strips metadata. A full audit trail supports the privilege log.
Full use case: Legal Use Cases.
When This Approach Has Limits
Replacing highlight-and-hope with true redaction removes the most common sanction trigger, but a tool does not make the underlying review judgment for you, and several risks survive the switch.
The tool removes what you mark — deciding what to mark is still yours. Automated detection can suggest privileged or personal spans, but it does not know your matter, your protective order, or what a particular passage implies in context. A name the model misses, or a privileged inference spread across sentences, remains exposed unless a human reviewer catches it. The copy-paste verification step is not optional.
Detection accuracy varies with the document. Scanned exhibits, handwriting, low-quality OCR, and unusual formatting lower recall. The cleaner the source text, the more the tool catches; the messier it is, the more manual review you need before filing.
Reversible tokenization shifts the risk to key custody. Using reversible encryption to satisfy later discovery orders means the original is recoverable — by anyone with the key. That is the right trade-off for evidence you may have to produce, but it makes key storage and access control part of your privilege-protection duty, not an afterthought.
Metadata and downstream copies can still leak. Even perfect text redaction does not help if document properties, tracked changes, or an earlier draft on a shared drive still carry the original. Treat redaction as one step in a production checklist that also covers metadata, versioning, and where copies live.