A security checklist for lawyers on sanitizing PDF metadata before sharing discovery

Key Takeaways

• PDF metadata can expose hidden client information—author names, document paths, editing software, timestamps—even after you have "removed" sensitive content from the visible text.
• Lawyers handling discovery must treat metadata sanitization as a distinct step in their workflow, separate from redaction and encryption.
• Browser-based, local-processing tools that use a Bring Your Own Key (BYO) model can process confidential PDFs without uploading the file to a third-party server, reducing exposure risk [K1].
• A reliable sanitization process combines three layers: automated metadata stripping, visual inspection of document properties, and verification of the output file.
• The cost of a metadata leak often exceeds the cost of the underlying dispute—reputation, privilege, and client trust are at stake.

1. Introduction

Discovery is where confidentiality goes to die—if you are not careful. When lawyers exchange PDFs during discovery, they focus on the visible content: the redactions, the text, the exhibits. But every PDF carries a hidden layer of information. The document's metadata—the author name, the software version, the original file path, the creation and modification timestamps, and sometimes even the names of previous reviewers—travels silently with the file [K1].

The problem is not hypothetical. Law firms have faced sanctions, privilege waivers, and professional embarrassment because a PDF sent to opposing counsel still contained the editing history that revealed internal strategy. The pain point is real: you cannot see metadata by looking at the document on screen, and most lawyers have never been trained to check for it.

This article provides a practical security checklist for sanitizing PDF metadata before sharing discovery materials. It explains what metadata exists, why it leaks, how to remove it reliably, and how to verify the result. It also introduces a privacy-preserving approach to AI-assisted PDF processing that keeps both the file and your API key local to your browser [K1]. By the end, you will have a repeatable process—not just a list of warnings.

2. Where Metadata Hides: Understanding the Risk

Core conclusion: PDF metadata is not a single field; it is a collection of embedded data blocks that can expose more than you expect. The safest assumption is that your PDF contains metadata until you have verified otherwise.

Let us break down what lives inside a typical PDF:

The risk is not just technical. When you share a PDF in discovery, the receiving party can extract this metadata with free tools in seconds. If your internal reviewer notes a concern in an annotation, or if the document path reveals that it was stored in a folder called "Settlement Strategy," you have just handed opposing counsel a roadmap.

Scenario: Your associate prepares a PDF of a contract for production. She opens the original Word file, converts it to PDF, and uploads it to your document management system. Unbeknownst to her, Word has embedded the author's full name, the original server path, and the names of three previous editors in the document properties. You send the PDF to opposing counsel. They run a basic metadata check and learn the document was reviewed by a partner who specializes in high-risk litigation—and that the file was last saved to a folder named "Privileged_Review." The privilege log dispute begins immediately.

Recommendation: Treat metadata sanitization as a separate quality gate. Do not assume that converting a file to PDF removes hidden information—it often carries it forward. Make a habit of checking metadata before the file leaves your firm's environment, not after.

3. The Sanitization Process: A Step-by-Step Checklist

Core conclusion: A reliable sanitization workflow has three stages: strip, verify, and protect. Skipping any stage leaves a gap in your security posture.

Stage 1: Strip Metadata from the Source

Start as early as possible in the document lifecycle. The most effective way to avoid metadata leaks is to remove them before they ever enter the PDF. If you generate the PDF from Microsoft Word, use the built-in Document Inspector (File > Info > Check for Issues > Inspect Document). It flags hidden properties, comments, and personal information before you convert the file.

For existing PDFs, use a dedicated PDF editor that offers a "Remove Metadata" or "Sanitize Document" function. These tools clear the document properties, flatten annotations, and remove hidden layers. The exact menu names vary by tool, but the approach is the same: you are rewriting the PDF without the embedded data.

If you are working with a PDF that contains comments, sticky notes, or highlight annotations, you must flatten them or remove them explicitly. Simply deleting them from view is not enough—annotations can remain in the file structure even when they are not visible on screen.

Stage 2: Verify the Output

After stripping, check the resulting file. Do not trust the tool's confirmation message. A simple verification workflow:

1. Open the cleaned PDF in your viewer.
2. Look at the document properties (usually File > Properties, or Ctrl+D/Cmd+D).
3. Confirm that the author is blank, the title is generic (or blank), and the producer field does not leak your firm's software versions.
4. Check for any remaining annotations or comments in the comments panel.
5. Use a separate metadata checker (many free tools exist) to see if any data remains.

Caveat: Some PDF viewers display metadata differently. A property that appears blank in one viewer may still be present in the file's raw structure. For high-stakes documents, use a command-line tool like exiftool for a thorough scan. It will show you every metadata field the file actually contains.

Stage 3: Protect During Processing

The stripping step removes metadata from the file, but you still need to process the document—maybe you want to summarize it with an AI tool, translate sections, or compare versions. This is where a second leak risk appears: the file may leave your control during processing.

Most cloud-based PDF tools upload your document to a server for processing. Even if the server is reputable, you have now handed over a client's confidential material to a third party. For discovery documents, that may be a privilege violation in itself.

This is where the BYO Key (Bring Your Own Key) model becomes relevant [K1]. Tools built on this model (such as OctopusPDF's Summarize and Translate features) process the file entirely in your browser. The PDF bytes are parsed locally with pdf.js—they never upload to a remote server [K1]. You supply your own OpenAI-compatible API key, which stays in your browser's localStorage and is never received by the tool's backend [K1]. The browser connects directly to the AI endpoint you choose, with no server-in-the-middle [K1].

What this means for lawyers: You can use AI to summarize a long contract or translate a foreign-language exhibit without sending the document to a cloud server. The service provider never sees the file or your API key [K1]. For discovery materials that are subject to privilege or confidentiality obligations, this is a meaningful security advantage.

Recommendation: Build a workflow that separates concerns. Use your firm's approved desktop tool for the actual metadata stripping. Use a browser-based BYO tool for AI-assisted tasks that require sending the text to an AI model—and ensure that the text goes directly from your browser to the model, not through an intermediary [K1]. Never upload a discovery PDF to a generic cloud converter without checking whether the file will be stored on their servers.

4. Handling Common Failure Points

Core conclusion: Metadata leaks are rarely caused by a single catastrophic error; they come from overlooked routine actions. The most common failure points have identifiable patterns.

Failure Point 1: Redaction ≠ Sanitization

Lawyers often confuse redaction with metadata removal. Redaction hides text visually—it may still exist in the PDF's underlying content stream or in hidden layers. A redacted PDF can still contain the original text if the redaction is not implemented properly.

Boundary condition: If you need to redact text, verify that the redaction tool performs a "burn-in" of the redaction (removing the underlying text) rather than just drawing a black box over it. Test the output by copying and pasting the visible area—if any text comes through, the redaction is inadequate.

Failure Point 2: Cloud Converters Leave Traces

Free online PDF converters are tempting, but they often store uploaded files on their own servers. Some keep them for hours or days. A discovery document uploaded to such a service is no longer under your control.

Practical rule: If you cannot confirm in writing that a PDF service deletes files immediately after processing and does not store them, do not use it for discovery materials. The BYO Key model sidesteps this problem entirely because the file never uploads in the first place [K1].

Failure Point 3: Metadata Reappears During Collaboration

Suppose you sanitize a PDF, send it to a colleague for review, and she opens it in an editor that adds her own metadata—her name, the new modification timestamp, and her software version. When you send that file to opposing counsel, you are back to square one.

Recommendation: Sanitize the file as the last step before sending. Do not sanitize, then edit, then send. If a document must go through review, sanitize it again at the end of the review cycle and verify once more.

5. Comparing Sanitization Approaches

Different scenarios call for different levels of care. This table compares the main approaches for handling discovery PDFs:

A recommended sequence: use a desktop sanitize function to clear metadata, verify with a command-line tool for high-stakes documents, and use a BYO-based browser tool for AI tasks that would otherwise require uploading the file [K1].

6. FAQ

Q1. Does converting a PDF to another format (e.g., DOCX) remove the metadata?

Converting a PDF to DOCX and back again may remove some metadata, but it can also create new metadata—including the conversion software's name and the conversion timestamp. It is not a reliable sanitization method. Use a proper sanitization tool instead.

Q2. How do I check what metadata is in a PDF?

In most PDF viewers, open the file and go to File > Properties (or press Ctrl+D / Cmd+D). This shows basic fields like author, title, subject, and producer. For a deeper scan, use a command-line tool like exiftool, which lists every metadata field embedded in the file.

Q3. Why does the BYO Key model help with privacy if I still send the file to an AI endpoint?

The key distinction is who receives the file. In a typical cloud tool, the file goes to the tool's server, which then sends it to an AI model. In a BYO Key model, the file never reaches the tool's server—the browser connects directly to the AI endpoint you choose [K1]. The service provider (the tool vendor) never sees the file or your API key [K1]. You still send the document's content to the AI model you selected, but you control which provider that is, and no intermediary stores a copy.

Q4. Is it enough to clear the author field and the title?

No. Author and title are the most visible fields, but the file path, producer, timestamps, and annotations can be equally revealing. Clear all document properties, flatten annotations, and verify the output with a separate checker.

7. Conclusion

Sanitizing PDF metadata is not a nice-to-have; it is a professional obligation when you share discovery materials. The risks are concrete: hidden author names, file paths, editing histories, and annotations can leak privileged information that no redaction tool will catch.

Follow a three-stage workflow: strip metadata from the source, verify the output with a separate tool, and protect the document during any further processing. For AI-assisted tasks—summarizing a long contract, translating an exhibit, or extracting key clauses—prefer tools that process files locally in the browser and use a Bring Your Own Key model, so the document never uploads to a third-party server [K1]. This keeps your confidential files in your control from start to finish.

The next time you prepare a PDF for discovery, take the extra three minutes to sanitize and verify. The cost of a leak—a waived privilege, a sanctioned firm, a client's lost trust—is far higher than the effort required to prevent it.