Famous PDF redaction failures — and the 30-second check that prevents them

Almost every public redaction failure shares one root cause: the tool drew a shape over the text instead of removing the text. A black rectangle is an annotation — a graphic layered on top. The words underneath are untouched, so select-all, copy-paste, search, and text extraction all still reveal them.

The second most common cause is metadata: the body is genuinely redacted, but the author name, document title, or an embedded prior version still names the person the redaction was meant to protect.

Both failures look identical to a correct redaction on screen. That's exactly why they slip through review — and why a mechanical check beats a visual one.

2005 — the Calipari report. The U.S. military published its investigation into the shooting of Italian intelligence officer Nicola Calipari as a redacted PDF. The blacked-out passages — classified operational details and names — could be recovered with select-all and copy-paste. The incident prompted the NSA to publish official guidance that same year warning that drawing black boxes over text does not remove it.

2009 — the TSA screening manual. The TSA posted its airport Standard Operating Procedures to a federal contracting website with the sensitive parts covered by black boxes. The boxes were trivially removable, exposing screening secrets — including sample law-enforcement credentials and the fact that only a fraction of checked bags were hand-searched. Congress opened an inquiry.

2011 — the UK nuclear submarine report. Responding to a freedom-of-information request, the Ministry of Defence published a report on Royal Navy submarine safety in which the 'redacted' text had simply been set on a black background. Copy-paste revealed assessments of how a structural failure could lead to a reactor meltdown aboard a nuclear submarine.

2019 — the Manafort filing. Paul Manafort's own defense lawyers filed a court document with black-box redactions over live text. Reporters pasted the contents into a text editor within minutes, revealing that Manafort had shared 2016 presidential campaign polling data with Konstantin Kilimnik — one of the biggest stories of the Mueller investigation, published by the defense itself.

2020 — the Maxwell deposition. This one is the cautionary counterexample: the redacted text really was removed. But the released transcript kept its alphabetized index, and Slate journalists used each redacted entry's position between known words to reconstruct the hidden names — including Bill Clinton, Alan Dershowitz, and Prince Andrew — within hours. Redaction isn't just removing the words; it's removing every path back to them.

Different institutions, fifteen years apart, same result: covered is not removed. The lesson isn't 'those people were careless.' It's that overlay-style redaction is a trap that produces a convincing-looking result while leaving the data intact. If your workflow can draw a box without destroying what's under it, you are one distracted afternoon away from the same headline.

Use a tool that removes the underlying text and image data from each redacted region — not one that draws over it — and then run the three checks above before you share the file.

PDF Redactly destroys the underlying data inside every redacted region (copy-paste and search reveal nothing), strips metadata, and does it entirely in your browser, so the document is never uploaded in the first place. But whatever tool you use, run the select-all, search, and metadata checks: 30 seconds of verification has prevented every failure on this list.