River
Y CombinatorBacked by Y Combinator

LegalFree

Litigation Redaction Log and Audit Check

River logs every redaction by Bates with its basis, then searches the rest of the production for the same value sitting in the clear.

Start here

A redaction log is a spreadsheet everyone knows how to fill in. Bates, page, what was covered, why, who did it, when. River builds that as it reads, one row per mark, with the basis written against the paragraph of the protective order or the rule that actually permits it rather than a one-word category. That part is table stakes, and it is also the part that gets audited when the other side moves to compel.

Then it inverts the log. Every distinct value behind those marks becomes a search term, and the search runs against the text of the entire production, including the pages carrying no redactions at all. A supplier name blacked out in one email and printed in the next attachment is not a close call. It is the same information, protected on one page and produced on another, and no reviewer looking at documents one at a time can see it.

Two more checks run in the same pass. Every mark is tested for whether it removed anything, because a box drawn over text and never applied leaves the words underneath. And the arithmetic around each redacted cell is worked, since a price hidden beside a quantity and a total is recoverable by division. The set this runs against usually arrives through a production that has to load before anyone reads it, and the privilege calls behind it belong in a log written to the elements of the claim.

A black rectangle is not a redaction

The PDF specification splits the job in two. A redact annotation identifies the region, and content removal is a separate step the reader performs afterwards. Marking content that way is non-destructive by design, so a review can happen before anything is deleted. Ship the file between those two steps and the box is decoration over live text. Pellworth's vendor shipped 88 such marks across 14 documents, every one of them returning the words underneath to a copy and paste.

When something privileged goes out, the rules give a route back that is not automatic. An inadvertent disclosure does not waive the privilege where the holder took reasonable steps to prevent it and promptly took reasonable steps to rectify the error. The notified party must then return, sequester or destroy the material and stop using it until the claim is resolved. Reasonable steps is the phrase that decides it, and a dated sweep is evidence of them.

Over-redaction has its own cost. What may be blacked out of a court filing is a short, closed list. The last four digits of a social-security or taxpayer number, the year of a birth date, a minor's initials, and the last four of a financial-account number. Anything beyond that in a filing needs an order for good cause requiring redaction of additional information. In a production it needs the protective order to reach it, which is why the basis column names a paragraph rather than a category.

How it works

  1. Hand over the set

    The documents, the redactions as applied, the protective order and any rule the court added.

  2. The log builds

    One row per mark, carrying the Bates, the page, the basis paragraph and the reviewer.

  3. The sweep runs

    Each protected value is searched across every page of the production, redacted or not.

  4. The exceptions come back

    Every inconsistency, live text layer and recoverable figure, each with the Bates numbers involved.

What you get

  • Every redaction logged by Bates and page, with the basis and the reviewer
  • The whole production searched for every value the log says is protected
  • Each mark tested for whether it removed the text or only covered it
  • Arithmetic worked around redacted cells, because a total and a quantity give back a price
  • Families checked together, so a clean attachment under a redacted parent is caught
  • Every entry whose logged basis reaches further than the protective order does, flagged

Common questions

Does it apply the redactions, or only check them?

It checks and logs. Applying redactions is a job for the review platform that holds the documents, and swapping that out mid-production is how Bates numbers stop matching. What comes back is the log you produce and two exception reports keyed to the same Bates numbers, so the fixes go back into the tool your team already uses.

The production already went out. Is this still worth running?

Yes, and sooner rather than later. Where privileged material slipped, the route back turns on having taken reasonable steps to prevent the disclosure and having promptly moved to rectify it. A dated sweep is both. It also tells you which specific Bates numbers to name in the notice, which is the difference between a clawback and an argument about scope.

How does it know what counts as the same value?

The log defines the terms. Every mark records what sits under it, so a supplier name, an account number or a price becomes a search string with its variants: the abbreviation, the entity suffix, the number with and without separators. Near matches are reported separately from exact ones, because a partial hit is a judgment call rather than a finding.

Is a redaction log the same as a privilege log?

No, and running them together is how both get thin. A privilege log covers documents withheld in full and has to describe each claim well enough to be assessed. A redaction log covers material blacked out inside documents that were produced. Where a mark rests on privilege it belongs on both, which is why proving each claim against its own elements stays a separate job.

What about scanned pages and images?

They get read the same way, and they are where the worst failures live. A black box drawn over a scan hides nothing if the page is text-searchable underneath, and an image clipped rather than cut still carries the original data in some formats. Pages that cannot be read at all are listed rather than passed, because an unreadable page is not a checked page.

Can it prioritise which documents get checked first?

Yes, and the order matters once the production is large. Documents already carrying weight get swept first: the ones on the exhibit list, the ones a witness has been asked about, and the ones sitting at contested dates in the chronology the production established. The remainder runs behind them rather than holding up the pages that matter.

What do I hand over at the end?

The log itself, in the column order your protective order or the local rule expects, plus the exception report the log cannot show. That second sheet is internal. It names each value protected in one place and clear in another, with both Bates numbers, so the fix is a re-production of specific pages rather than a rolling response with a deadline of its own.

Litigation Redaction Log and Audit Check

Fill in the form and your workspace opens with the work already underway.